What does the research say?
In these studies, agitation and behavioral symptoms were measured with scales filled in by a caregiver or nurse who knew the patient (Cohen-Mansfield Agitation Inventory, Neuropsychiatric Inventory, Pittsburgh Agitation Scale). Most patients were very old and had moderate or severe dementia, and many were already taking antidepressants, antipsychotics or sedatives. Behavioral symptoms often went down in people taking an inactive pill (placebo) too: for example, in the CBD-rich oil study, the agitation score dropped by at least 4 points in 30% of those on placebo as well. This is why it is hard to credit the product for improvements seen in studies without a comparison group.
There are only two small controlled studies with pure CBD. In 30 people with dementia caused by blood vessel problems (vascular dementia), 300 mg of CBD a day for four weeks reduced behavioral symptoms slightly more than placebo; the difference was right at the edge of statistical significance, there was no difference in thinking skills or daily functioning, and three of the eight authors were among the inventors on two CBD-related patents. In a six-week pilot study with 15 people with Alzheimer’s disease (up to 600 mg a day), behavioral scores improved numerically more in the CBD group; but the study was not designed to test whether it works, and 5 of the 8 people taking CBD had dizziness. In a six-month study in Greece comparing CBD drops with usual treatment in 20 people, improvement was reported in the CBD group; however, patients were not assigned to groups by chance, and the abstract does not say whether the product contained THC. For pure CBD the strength of evidence is very weak; as of October 2026, several controlled studies with pure CBD are under way.
Studies with products rich in CBD but also containing THC gave different results. In a 16-week study with 60 patients in Israel (an average of about 528 mg of CBD and 22 mg of THC a day), the share of people whose agitation clearly went down was 60% in the treatment group and 30% in the placebo group; however, with the statistical correction the authors had set in advance, this difference was borderline (p = 0.06). Also, all 8 patients who dropped out were in the treatment group, complaints of memory loss and hallucinations (seeing or hearing things that are not there) were numerically more common in the treatment group, and the study was funded by the company that owns the product. In Switzerland, in 25 nursing home residents with severe dementia, an oil containing THC and CBD in a 1:2 ratio did not reduce agitation differently from placebo over eight weeks; only the use of sedatives given as needed was lower. In a pilot study in England with 29 nursing home residents, in which a mouth spray containing THC and CBD was tried for four weeks, and in a study in Australia with a THC-dominant oil in 21 people, no clear difference in behavioral measures was seen either (the Australian study reported a reduction in agitation only at the end of treatment). In a six-month study in Brazil with an extract containing very low doses of THC and CBD in 28 Alzheimer’s patients, the score on a general mental test was better in the treatment group; there was no difference in daily functioning or behavior, and fewer patients were enrolled than planned. Because these products also contain THC, their results count toward the evidence for THC-containing products; the strength of evidence is weak.
In studies with THC-containing medicines, the picture is also mixed. In two studies of 50 and 22 people in the Netherlands with a low-dose THC tablet (1.5–4.5 mg a day), behavioral symptoms did not go down differently from placebo. In a three-week study at five centers in the US with 75 Alzheimer’s patients, dronabinol, a synthetic THC (up to 10 mg a day), was better than placebo on one of the two main scales measuring agitation, while on the other the difference was not significant; there was no difference in sleep, daily activities, thinking or the other measures. In a crossover study in Canada (a design in which patients take the drug and the placebo in turn) with nabilone, a synthetic drug similar to THC, in 39 patients with moderate to severe Alzheimer’s disease, agitation went down more than with placebo; however, sleepiness was clearly more common during the drug period (45% vs. 16%), and on a thinking test used for severe dementia the result favored placebo. In a small crossover study in 1997 with 15 Alzheimer’s patients who refused to eat, weight went up more and disturbed behavior decreased during the dronabinol period. For THC-containing products the strength of evidence is weak.
Analyses pooling these studies did not reach the same conclusion. A Cochrane review prepared by independent experts (2021; 4 studies, 126 people) wrote that it cannot be said with confidence whether cannabinoids are helpful or harmful in dementia, and that any benefit may be too small to make a difference in daily life. Of three meta-analyses published in 2026, one (9 studies, 334 people) found no effect on agitation, behavioral symptoms or thinking; one (10 studies, 328 people) found a reduction in agitation, but the difference was no longer significant when studies at high risk of bias were removed; the third, limited to Alzheimer’s (7 studies, 221 people), found a small to moderate reduction in behavioral scores. All three agreed that sleepiness was about twice as common in people taking these products; according to Cochrane, the certainty of the evidence is low. Canada’s guideline on mental health in older adults (2025) lists synthetic cannabinoids such as nabilone among the last options for severe agitation that has not responded to other medicines, as a conditional recommendation based on low-quality evidence; the 2026 assessment by Canada’s Drug Agency, however, found the benefit of nabilone uncertain because it rests on a single small study; the Brazilian Academy of Neurology (2021) does not recommend their use in dementia. Overall, the strength of evidence is weak.
There is no large study designed specifically for sleep, appetite or caregiver burden. Sleep: in the study of the CBD-rich oil that also contains THC, sleep problems went down more in the treatment group (daytime sleepiness was also more common in the same group); in the dronabinol study, there was no difference in sleep. Appetite and weight: the most direct data come from the 1997 dronabinol study with 15 patients who refused to eat. Caregivers: in the nabilone study, the caregiver distress score was slightly lower during the drug period. In these areas the strength of evidence is very weak.
Studies without a comparison group report more positive results. In Colombia, in 60 Alzheimer’s patients using a CBD-rich oil (containing a small amount of THC), symptom scores fell by about half in three months; in Geneva, in 19 nursing home residents with severe dementia given an oil containing THC and CBD, improvement and a reduction in other medicines were reported over follow-up of up to 13 months; in two chart reviews in the US (40 and 50 patients), improvement was noted with dronabinol or with THC products bought from retail outlets. Because patients in these studies were not assigned to groups by chance and there was no placebo group, it is not known how much of the improvement was due to the product and how much to expectation or to changes that happen on their own over time. The controlled study the Geneva team carried out later did not confirm the positive result of the early observations on the agitation measure. Studies of this kind are very weak evidence.
Claims that “CBD prevents Alzheimer’s” or “protects the brain” have not been shown in people: there is no human study with published results that used a cannabinoid to prevent or slow dementia; this idea comes from experiments in animals and in the laboratory. A study testing CBD in 236 people with mild cognitive impairment is expected to finish in 2029. Large observational studies point both ways: in one analysis of about 80,000 older adults in England, having used cannabis was not significantly linked to faster cognitive decline, and in US veterans cannabis use disorder was not significantly linked to dementia risk, and a genetic analysis showed no causal link; in the Canadian province of Ontario, however, middle-aged and older adults who went to the emergency room or were admitted to hospital because of cannabis were diagnosed with dementia more often over the next five years. These studies are observational, which means they cannot show cause and effect on their own. For prevention there is no evidence in humans (insufficient); the evidence on a link between cannabis use and dementia risk is also very weak.
Strength of evidence: ●○○○ Very weak evidence. Finding: Not studied enough. What does this mean?
Ongoing research on this topic
Trials testing CBD for this topic on ClinicalTrials.gov: 5 under way (3 recruiting); results are awaited from 2 that ended in the last two years. A trial being under way does not show that CBD works.
- NCT05822362: Phase 2, recruiting, target 236 participants
- NCT05644262: Phase 2, completed, results awaited, 120 participants
- NCT06570928: Phase 2, completed, results awaited, 72 participants
Possible side effects and drug interactions
The most consistent side effect in dementia studies was sleepiness and drowsiness: in analyses pooling the studies it was about twice as common in people taking these products; with nabilone it was 45% vs. 16%. With the CBD-rich oil that reached an average of 22 mg of THC a day, memory loss, hallucinations and confusion were reported numerically more often. In the small study in which pure CBD was tried at up to 600 mg a day, dizziness was common and liver enzymes rose. With the low-dose THC tablet, standing balance with the eyes closed got slightly worse, but there were no falls.
These side effects matter because falls, daytime sleepiness and confusion are already common in older adults with dementia. In the dronabinol study, two serious events (one episode of sudden confusion and one broken arm bone) occurred in the treatment group and none in the placebo group; the number of falls was similar in the two groups. In an analysis of 46 controlled studies in adults over 50, THC-containing products increased side effects, while pure CBD did not.
People with dementia often take many medicines. In the long-term follow-up in Geneva, an oil containing THC and CBD slightly reduced the activity of two liver enzymes (CYP1A2 and CYP2C19) that help break down certain medicines. This is why it is important to review all the medicines being taken together with the doctor and pharmacist.
Interactions with medicines used for this condition
Interactions reported for these medicines in peer-reviewed publications and official prescribing information. A medicine that is not listed can still interact with CBD; tell your doctor about every medicine you take.
| Medicine | Importance | Warning | Source |
|---|---|---|---|
| Other benzodiazepines (lorazepam, etc.), Sleeping pills and other sedatives (zolpidem, etc.) | Caution | Benzodiazepines, sleep medicines and sedatives: Drowsiness and sedation may increase together with CBD. Talk to your doctor before using them together. | DailyMed Epidiolex label 7.4, EMA Epidyolex SmPC 4.4 |
| Other antipsychotics (haloperidol, etc.) (Norodol) | Information | Antipsychotics: No published study has measured antipsychotic blood levels with CBD. In schizophrenia studies, CBD (600–1000 mg a day) added for 6 weeks to the antipsychotic being taken was generally well tolerated; in one study, drowsiness was more common in those taking CBD. Drowsiness may increase when they are taken together; tell your doctor. | PMID 29241357, PMID 29619533, FDA Epidiolex label §7.4 (05/2026) |
To check interactions with the medicines you take: CBD drug interaction checker.
How was CBD used in studies?
The doses used in the studies varied widely by product. Pure CBD was given at 300 mg a day (study dose; not a personal recommendation) or started at 200 mg a day and increased up to 600 mg (study dose; not a personal recommendation). With the CBD-rich oil that also contains THC, the dose was increased drop by drop and reached an average of about 528 mg of CBD and 22 mg of THC a day (study dose; not a personal recommendation); with the oil containing THC and CBD in a 1:2 ratio, the dose was raised over two weeks to at most 20 mg of THC and 40 mg of CBD a day (study dose; not a personal recommendation). Dronabinol was started at 5 mg a day and raised to 10 mg (study dose; not a personal recommendation), and nabilone was raised to 1–2 mg a day (study dose; not a personal recommendation); the low-dose THC tablets were 1.5–4.5 mg a day (study dose; not a personal recommendation). Most studies lasted between 2 weeks and 6 months, and the medicines were given under doctor and nurse supervision.
Study details (for health professionals)
Controlled studies with cannabinoids for behavioral symptoms in dementia (agitation, aggression) are small (15–75 people), short (2 weeks to 6 months) and contradict each other. There are only two RCTs with pure CBD: in vascular dementia, a 30-person study with 300 mg of CBD a day showing borderline superiority over placebo on the NPI and BPRS (three of the eight authors are inventors on CBD-related patents), and in Alzheimer’s, a 15-person feasibility study with up to 600 mg of CBD a day that did not test efficacy (dizziness 63%). With CBD-rich products that also contain THC, the results point both ways: in a 60-person, manufacturer-supported RCT with an oil reaching an average of 528 mg of CBD and 22 mg of THC a day, agitation response was 60% vs. 30% (p = 0.06 with the continuity correction prespecified in the statistical plan; all dropouts in the treatment group; memory loss and hallucinations numerically more common); with a THC:CBD 1:2 oil (25 people, crossover), a mouth spray containing THC and CBD (29 people, feasibility) and a THC:CBD 3:2 oil (21 people, crossover), no difference on behavioral measures (in the last, a reduction in agitation only at the end of treatment); in a 28-person study with a very low-dose THC-CBD extract, only the MMSE favored the drug. THC and nabilone: two RCTs with a low-dose THC tablet were negative; in a 75-person multicenter RCT with dronabinol, only one of two co-primary measures was positive; in a 39-person crossover RCT with nabilone, agitation went down and sedation went up (45% vs. 16%); a very small 1997 dronabinol study with 15 Alzheimer’s patients who refused to eat reported positive signs for weight and behavior. Studies without a comparison group (19–60 people) reported marked improvement; these designs cannot separate out the effect of expectation and of improvement that happens on its own over time, and the same team’s later controlled study (Geneva) did not confirm the positive observation. Cochrane 2021 (4 RCTs, 126 people) wrote that it cannot be said with confidence whether there is benefit or harm, with low certainty for behavioral symptoms and very low certainty for cognition. Three meta-analyses in 2026 reached different conclusions: one found no effect (9 RCTs; moderate certainty for behavior), one found a significant reduction in agitation only when studies at high risk of bias were included (10 RCTs), and the third, limited to Alzheimer’s, found a small to moderate benefit (6 RCTs and 1 open-label study); in all three, sleepiness was about twice as common (RR 2.03–2.25). The 2025 guideline of the Canadian Coalition for Seniors’ Mental Health gives synthetic cannabinoids a conditional place, based on low-quality evidence, in severe agitation that has not responded to other medicines (the guideline authors include two investigators from the nabilone study in the table); the 2026 assessment by Canada’s Drug Agency found the nabilone evidence uncertain; the Brazilian Academy of Neurology (2021) does not recommend use in dementia; cannabinoids are not included in the Korean (2025) and Taiwanese (2026) guidelines or in the recommendations of Canada’s 5th consensus conference on dementia (2020). There is no human study with published results that used a cannabinoid to prevent or slow dementia; in large observational studies, having used cannabis was not significantly linked to cognitive decline, and cannabis use disorder was not significantly linked to dementia risk (UK Biobank, US veterans; no causality in Mendelian randomization), but middle-aged and older adults who received emergency care because of cannabis were diagnosed with dementia more often over the next 5 years (Ontario; observational). October 2026 search (PubMed, Cochrane, ClinicalTrials.gov): on this topic we found 12 randomized controlled studies with cannabinoids (the largest 75 people; only 2 small studies with pure CBD), 1 non-randomized controlled study, 5 studies without a comparison group, 2 large observational studies looking at dementia risk, 6 systematic reviews (2 of them Cochrane, 1 a safety review in older adults), 1 guideline, 1 position paper and 1 health technology assessment; all are in the table below. 3 controlled studies that were completed but whose results have not been released, and ongoing trials, are listed under “unverified”; open-label studies with fewer than 15 people, case reports, a crossover study completed with only 2 patients, and a 24-person study that did not report the dronabinol result separately were not included in the table.
Evidence assessment: Very weak: there are only two small RCTs with pure CBD (30 and 15 people, 4–6 weeks); one is borderline positive and three of its authors are inventors on CBD-related patents, the other is a feasibility study that did not test efficacy; there is one non-randomized 20-person comparison with CBD drops of unknown THC content. Results for products rich in CBD but also containing THC were assessed under the THC evidence
CBD alone
There are only two small controlled studies with pure CBD: in 30 people with vascular dementia, 300 mg of CBD a day for four weeks reduced behavioral symptoms slightly more than placebo (right at the edge of statistical significance; three of the eight authors among the inventors on two CBD-related patents); a six-week pilot study with 15 Alzheimer’s patients (up to 600 mg a day) was not designed to test whether it works, and dizziness was common. A 20-person study that did not assign people to groups by chance reported improvement with CBD drops; whether the product contained THC is not stated. As of October 2026, several controlled studies with pure CBD are under way. The strength of evidence is very weak.
Products containing THC
The results of 10 small controlled studies with THC-containing products and with the THC-like synthetic drug nabilone are conflicting. Two studies with a low-dose THC tablet (50 and 22 people) found no difference; in a 75-person multicenter study with dronabinol, benefit was seen on only one of the two main scales (there were differences between the groups at the start); in a 39-person crossover study with nabilone, agitation went down and sleepiness went up. In a 60-person study with an oil rich in CBD but reaching an average of 22 mg of THC a day, agitation went down, but with the statistical correction the authors had set in advance the difference on the main measure was borderline (p = 0.06), and everyone who dropped out was in the treatment group; four studies with oils, sprays and extracts containing THC and CBD found no clear difference in agitation (in one, a reduction only at the end of treatment). The Cochrane review rated the certainty of the evidence low or very low; three meta-analyses in 2026 reached different conclusions, and in all of them sleepiness was about twice as common. Canada’s 2025 guideline lists synthetic cannabinoids such as nabilone among the last options for severe agitation that has not responded to other treatments, as a conditional recommendation based on low-quality evidence; the 2026 assessment by Canada’s Drug Agency, however, found the benefit of nabilone uncertain because it rests on a single small study.
Evidence: Weak: 10 small, short controlled studies with THC-containing products and nabilone (the largest 75 people, 3 weeks), with conflicting results; certainty low to very low according to Cochrane; the most consistent finding is that sleepiness was about twice as common
Dose range in studies
Pure CBD 300 mg/day and from 200 mg up to 600 mg/day; CBD-rich oil avg. ~528 mg CBD + 22 mg THC/day; THC:CBD 1:2 oil at most 20 mg THC + 40 mg CBD/day; THC:CBD mouth spray at most 10.8 mg THC + 10 mg CBD/day; THC:CBD 3:2 oil at most 50 mg THC/day (protocol); very low-dose extract 0.35 mg THC + 0.245 mg CBD/day; dronabinol 2.5–10 mg/day; low-dose THC tablet 1.5–4.5 mg/day; nabilone 0.5–2 mg/day
Starting dose and titration
In most studies, treatment started at a low dose and was raised to the target dose over 1–6 weeks (over 15 days with the THC:CBD 1:2 oil; with the CBD-rich oil, by adding one drop every two days over at most 6 weeks; with dronabinol, 5 mg in week 1, then 10 mg/day); these are research protocols
These are doses researchers gave under medical supervision; they are not a recommended dose for you. "mg/kg/day" means milligrams per kilogram of body weight per day.
| Study | Participants | Product | Dose and titration | Duration | Result | Source |
|---|---|---|---|---|---|---|
| Rosenberg 2026 (THC-AD) Phase 2, multicenter (5 centers), double-blind, placebo-controlled parallel RCT | 75 randomized (37/38); 63 completed Agitation due to Alzheimer’s disease (IPA criteria; NPI-C agitation/aggression ≥4), severe cognitive impairment; most were taking an antidepressant or antipsychotic (US) | Dronabinol (synthetic THC) capsule, added to existing treatment | Week 1: 2.5 mg twice a day (5 mg/day); weeks 2–3: 5 mg twice a day (10 mg/day) To 10 mg/day within 1 week | 3 weeks | On one co-primary outcome, the Pittsburgh Agitation Scale (PAS), the weekly decline was greater than with placebo (slope difference −0.74; 95% CI −1.33 to −0.15; p = 0.015); on the other co-primary outcome, NPI-C agitation/aggression, the difference was not significant (−1.26; 95% CI −2.73 to 0.21; p = 0.094); the abstract nonetheless reports a “clinically meaningful” reduction on both measures. No difference on any secondary outcome (sleep, activities of daily living, CMAI, cognition, intoxication, as-needed lorazepam or trazodone use). At baseline, PAS was higher in the dronabinol group (7.1 vs. 5.6; at week 3, 4.0 vs. 5.1) and fewer were taking antidepressants (7 of 37 people vs. 25 of 38; baseline table in the registry). Supported by the US National Institute on Aging (NIA); one of the 22 authors declares consulting or advisory board roles with cannabinoid companies (Charlotte's Web, Jazz, Syqe) Side effects: According to the abstract, no difference in side effects other than sleepiness. In the registry results: serious side effects 2 vs. 0 (delirium, broken arm bone), diarrhea 5 vs. 0, falls 3 vs. 5, sleepiness reported as a side effect 2 vs. 5; no difference in the number of side effects (event rate ratio 1.14; 95% CI 0.73–1.79; p = 0.57) | PMID 41350162, doi:10.1016/j.jagp.2025.10.011, NCT02792257, May 9, 2025, 84 participants in the registry, 75 randomized in the publication), PMID 34629131, PMID 41350163 |
| Hermush 2022 Single-center, double-blind, placebo-controlled parallel RCT (2:1) | 60 started treatment (40/20; 64 randomized, 4 withdrew immediately); 52 completed Major neurocognitive disorder with behavioral disturbance (all types of dementia; NPI-NH agitation ≥3, MMSE <26), aged ≥60; mostly at home under 24-hour supervision (Israel) | Avidekel: broad-spectrum cannabis oil, sublingual drops (30% CBD, 1% THC; 295 mg CBD + 12.5 mg THC per mL; 11.8 mg CBD + 0.5 mg THC per drop; also CBC, CBG, CBDV; contains THC) | Avg. 527.5 mg CBD + 22.3 mg THC/day (3 times a day) Started at 1 drop per dose, 3 times a day; increased by 1 drop per dose every two days up to at most 21 drops/dose; at most 6 weeks | 16 weeks (dose adjustment for up to 6 weeks, then fixed dose) | On the primary outcome (≥4-point drop in CMAI at week 16), 60% vs. 30% (χ² p = 0.03; p = 0.06 with the continuity correction prespecified in the statistical plan; in the 52 completers 69% vs. 30%, p = 0.01 with correction); ≥8-point drop 50% vs. 15% (p = 0.011); CMAI drop 10.7 vs. 2.5 points (p = 0.03). Data from dropouts were analyzed by carrying the last measurement forward (LOCF). On the NPI-NH, agitation/aggression and sleep disturbance were lower in the treatment group; no difference on the MMSE. The intention-to-treat analysis had 60% power; funded by the company that owns the product; two of the 9 authors are company employees (with share options), and one is the inventor on a patent for this use Side effects: All 8 patients who dropped out were in the treatment group (not because of side effects, according to the authors); memory loss 32% vs. 10% (p = 0.06), hallucinations 22% vs. 5% (p = 0.08), sleepiness 49% vs. 30% (p = 0.17), confusion 46% vs. 30%; serious events 9 vs. 4; 2 deaths in the treatment group (investigators did not find them related to the product) | PMID 36148467, doi:10.3389/fmed.2022.951889, NCT03328676 |
| van den Elsen 2015a Phase 2, multicenter, double-blind, placebo-controlled parallel RCT | 50 (24/26) Dementia with clinically relevant neuropsychiatric symptoms (Alzheimer’s, vascular or mixed; Netherlands) | THC tablet (Namisol) | 1.5 mg 3 times a day (4.5 mg/day) None (fixed dose) | 3 weeks | No difference from placebo on the primary outcome (NPI) (mean difference 3.2; 95% CI −3.6 to 10.0); symptoms went down in both groups. No difference in CMAI (4.6; −3.0 to 12.2), quality of life or activities of daily living either Side effects: Number of patients with mild to moderate side effects similar (16 vs. 14; p = 0.36); no effect on vital signs, weight or episodic memory | PMID 25972490, doi:10.1212/WNL.0000000000001675, NCT01608217, patients with pain or suspected pain (29 people) were also given paracetamol (acetaminophen), the drug was supplied by the manufacturer (Echo Pharmaceuticals), funding from the EU Regional Development Fund and the province of Gelderland |
| Herrmann 2019 Phase 2/3, double-blind, placebo-controlled crossover RCT (2 × 6 weeks, with a 1-week washout in between) | 39 randomized Moderate-to-severe Alzheimer’s disease (sMMSE ≤24) with agitation (NPI-NH agitation/aggression ≥3); long-term care facilities and geriatric psychiatry outpatient clinics; mean age 87, 77% men (Canada) | Nabilone (THC-like synthetic cannabinoid) capsule | Target 1–2 mg/day (mean 1.6 ± 0.5 mg) Per the protocol, from 0.5 mg to 2 mg | 14 weeks | Difference favoring nabilone on the primary outcome (CMAI) (b = −4.0; 95% CI −6.5 to −1.5; p = 0.003); NPI-NH total −4.6 (−7.5 to −1.6), caregiver distress −1.7 (−3.4 to −0.07; p = 0.041) and sMMSE +1.1 (0.1 to 2.0) favored nabilone; in the 25 people who completed the severe impairment battery, the result favored placebo (p = 0.003); improvement on the clinician’s global impression of change 47% vs. 23% (p = 0.09, not significant) Side effects: Sleepiness (sedation) 45% during the nabilone period, 16% during the placebo period (p = 0.02); difference in treatment-limiting sedation not significant (p = 0.22) | PMID 31182351, doi:10.1016/j.jagp.2019.05.002, NCT02351882, PMID 31338476, PMID 31322567, 24S-hydroxycholesterol), PMID 31547752, oxidative stress and cytokines), PMID 41530008, in a post hoc analysis, response was larger in those with marked pain, appetite and eating problems, or apathy), PMID 41220257, predicting response with blood markers, exploratory), PMID 31196620 |
| Pessoa 2026 Double-blind, placebo-controlled parallel RCT | 30 Vascular dementia (caused by blood vessel problems) with clinically significant behavioral and psychological symptoms, older adults (Brazil) | CBD, by mouth (the abstract says only “CBD 300 mg/day”; the product’s source and purity are not in the abstract, and the ReBEC registry entry could not be accessed) | 300 mg/day Not reported (in the abstract) | 4 weeks | Primary outcomes favored CBD: time × group interaction on the NPI F = 3.61, p = 0.05 (partial eta squared 0.11) and on the BPRS F = 4.02, p < 0.05 (partial eta squared 0.13); no difference in cognition (MMSE, IQCODE) or function (Katz, Lawton). Three of the 8 authors are co-inventors on patents for fluorinated CBD compounds and an oral drug formulation containing cannabinoids (the University of São Paulo licensed the first patent to a company and has an agreement with a pharmaceutical company to develop a synthetic CBD product); two of these three authors have received consulting fees from cannabinoid companies Side effects: Mild and similar in the two groups (UKU scale) | PMID 41277041, doi:10.1177/02698811251390974, ReBEC RBR-686kmpz |
| Cury 2025 Phase 2, double-blind, placebo-controlled parallel RCT | 29 randomized; 28 received treatment; 14 per group analyzed (precalculated target 37) Dementia due to Alzheimer’s disease, aged 60–80 (Brazil) | Balanced THC-CBD cannabis extract, by mouth (contains THC) | 0.350 mg THC + 0.245 mg CBD/day Not reported (in the abstract) | 26 weeks | On the primary outcome at week 26, the MMSE total score was higher in the cannabis group than with placebo (mixed model; size of the difference not in the abstract); no difference in secondary outcomes, including function and behavior. Two of the 13 authors are partners in a contract research company that runs cannabis studies Side effects: No significant difference in side effect frequency; according to a commentary, sleepiness about twice and suspiciousness (paranoia) five times as common (not significant; numbers could not be read) | PMID 41160460, doi:10.1177/13872877251389608, ReBEC U1111-1258-2058, PMID 42216668 |
| Albertyn 2025 (STAND) Double-blind, placebo-controlled feasibility RCT (nursing homes in England) | 29 randomized (target 60) Probable Alzheimer’s disease with clinically significant agitation; moderate-to-severe or severe stage, mostly frail nursing home residents | Nabiximols (Sativex) mouth spray; 2.7 mg THC + 2.5 mg CBD per spray (contains THC) | Target 4 sprays/day (10.8 mg THC + 10 mg CBD) 1 spray/day for the first 3 days, then up to at most 4 sprays/day; only half of participants followed the schedule fully | 4 weeks of treatment + 4 weeks of follow-up | Feasibility: the 60-person target was not reached (29 people in 7.5 months); retention and adherence 100%. The prespecified CMAI effect size threshold (d ≥ 0.3) was not met: d = 0.23 at week 4 (95% CI −0.2 to 0.7), 0.08 at week 8; adjusted CMAI difference 6.77 (−6.71 to 20.25). CMAI went down in both groups; at baseline, agitation was more severe in the nabiximols group (CMAI 95.5 vs. 71.5) Side effects: No side effects attributed to the product; mild side effects 4 vs. 2 people, serious events 1 vs. 1 person (unrelated to the product); no falls, deaths or withdrawals | PMID 40479610, doi:10.1093/ageing/afaf149, ISRCTN 7163562 (as written in the publication), drug and placebo from Jazz Pharmaceuticals (formerly GW Pharmaceuticals), funding from Alzheimer's Research UK, one of the 12 authors has served on advisory boards of cannabis companies (EmpowerPharm/SanteCannabis, NW PharmaTech) or received research support through their institution |
| Bianchi 2026 (MedCanDem) Multicenter, double-blind, placebo-controlled crossover RCT (2 × 8 weeks, 1-week washout) | 25 randomized; 19 completed both periods Severe dementia with behavioral or psychological symptoms; 5 long-term care facilities in Geneva; mean age 83; patients taking many medicines | Standardized cannabis oil, oral drops (Cannapharm; 1% THC, 2% CBD; THC:CBD 1:2; contains THC); placebo: hemp seed oil | After titration avg. 17.85 mg THC + 35.7 mg CBD/day (at most 20 mg THC + 40 mg CBD) Started at 2.5 mg THC + 5 mg CBD/day, increased every two days to at most 20 mg THC + 40 mg CBD/day within 15 days | 2 × 8 weeks | No difference on the primary outcome (CMAI) (crossover difference 5.0; SD 33.8; p = 0.53); no difference in NPI, pain (Doloplus) or the team’s global impression of change either. As-needed psychiatric medicine was given less during the active period (0.06 vs. 0.14 per person per day; p < 0.0001). At baseline there was a CMAI difference between the sequence groups, and the 1-week washout was too short (active breakdown products were found in some patients’ blood at the switch to the placebo period) Side effects: No serious drug-related side effects; sleepiness 3 vs. 2 events, falls 1 vs. 0; blood pressure changes beyond the protocol limit (±15%) 71 events in the active period and 28 in the placebo period (no statistical test reported) | PMID 42613673, doi:10.1093/ageing/afag239, NCT05432206, PMID 37293434, product and placebo provided free by the manufacturer, two authors are former board members of the Swiss Society for Cannabis in Medicine, the same team’s uncontrolled follow-up: Pautex 2022 |
| van den Elsen 2015b Phase 2, multicenter, double-blind, placebo-controlled, repeated crossover RCT (6 blocks × 2 weeks: in each block 3 days of THC and 3 days of placebo, with 4-day washouts in between) | 22 (20 completed) Dementia with clinically relevant neuropsychiatric symptoms; mean age 76; 2 hospitals (Netherlands) | THC tablet (Namisol) | Blocks 1–3: 0.75 mg twice a day; blocks 4–6: 1.5 mg twice a day (1.5–3 mg/day) Dose escalation (to 1.5 mg after 6 weeks) | 12 weeks | No difference from placebo on the primary outcome (NPI) (blocks 1–3: 1.8; 97.5% CI −2.1 to 5.8; blocks 4–6: −2.8; −7.4 to 1.8) Side effects: Side effect frequency similar in the two periods; 4 serious events unrelated to the drug. In substudies of the same trial: of 98 side effects reported in 10 patients, 6 were found to be related to THC, and at 1.5 mg body sway with the eyes closed increased (Ahmed 2015); in 18 patients, 1.5 mg twice a day increased sway with the eyes closed and stride length when walking, with no falls (van den Elsen 2017) | PMID 26560511, doi:10.1016/j.jagp.2015.07.011, NCT01302340, PMID 25752889, PMID 27624148, a secondary analysis of the same trial according to the Bahji 2022 review, could not be verified from the full text) |
| Timler 2023 Double-blind, placebo-controlled crossover RCT (2 × 6 weeks of treatment, with 2-week washouts) | 21 Older nursing home residents diagnosed with dementia; 77% women, mean age 85 (Australia) | CogniCann: THC:CBD 3:2 oil, sprayed into the mouth (per the protocol, 25 mg THC + 17 mg CBD per mL; contains THC) | Per the protocol, the highest dose the person tolerated, at most 50 mg THC (+34 mg CBD)/day; doses given are not in the abstract Per the protocol, started at 2.5 mg THC and increased over a few weeks | 18 weeks | No difference from placebo in behavior (NPI-Q, CMAI), quality of life or pain; only a reduction in agitation favoring the drug at the end of treatment. A later calculation showed 50 people would have been needed for the NPI (21 could be enrolled against a target of 50) Side effects: According to the abstract, few side effects; another team writes that total side effects were clearly more common in the active group in this study (Albertyn 2025); full text could not be read | PMID 37321847, doi:10.1111/ajag.13224, ANZCTR ACTRN12619000474156, PMID 32059690, funded by the product’s manufacturer, MGC Pharmaceuticals) |
| Velayudhan 2024 (CANBiS-AD) Phase 2a, single-center, double-blind, placebo-controlled parallel feasibility RCT (published as a letter to the editor) | 16 randomized; 15 received treatment (8/7) Possible or probable Alzheimer’s disease with behavioral symptoms (NPI ≥4), aged ≥55; 93% women, mean age 78 (London) | CBD capsule (200 mg; pure CBD) | From 200 mg/day to 600 mg/day (3 capsules a day) From 1 capsule a day to 3 capsules (details in the publication’s supplementary file, which could not be read) | 6 weeks | Primary outcomes were feasibility: treatment adherence 100%, retention 94%. Change from baseline in NPI-C total −29.9 ± 51.5 in the CBD group and −10.1 ± 38.2 in the placebo group; no statistical test was done because the study was not designed to test efficacy; delusions were more severe at baseline in the placebo group Side effects: 34 side effects in total; dizziness 63% vs. 0%, falls 25% vs. 43%; the most common drug-related side effects were mild, temporary dizziness, sleepiness and raised liver enzymes; no withdrawals or deaths; 1 serious fall in the placebo group after treatment ended; according to the same team’s later publication, dizziness attributed to treatment was 12.5% in the CBD group (1 of 8 people; Albertyn 2025) | PMID 38584388, doi:10.1017/S1041610224000516, EudraCT 2019-002106-52, study drug provided free by Beckley Canopy Therapeutics/Canopy Growth (investigator-initiated study), one of the 7 authors has served on advisory boards of cannabis companies (EmpowerPharm/SanteCannabis) or received speaker fees, funding from King's Health Partners, Psychiatry Research Trust and Rosetrees Trust |
| Volicer 1997 Placebo-controlled crossover study (2 × 6 weeks; controlled clinical trial in PubMed; randomized and double-blind according to reviews) | 15 (11 completed; 1 patient who died was included in the analysis) Patients with probable Alzheimer’s who refused to eat (hospital dementia unit, US) | Dronabinol (synthetic THC) | 2.5 mg in the morning and at noon (5 mg a day; according to the study table of Cochrane 2021; one review gives 2.5 mg a day) None (fixed dose) | 12 weeks | Body weight went up more during the dronabinol period than during the placebo period; the severity of disturbed behavior decreased, and in those who took dronabinol first this effect continued into the placebo period (abstract). Reviews report the CMAI result differently; the full text could not be read Side effects: Euphoria (an excessively elated mood), sleepiness and fatigue more common during the dronabinol period, treatment not stopped; study ended in 1 patient because of a major seizure and in 2 patients because of severe infection; 1 patient died of a heart attack | PMID 9309469, The only study included in the Krishnan 2009 Cochrane review (its data could not be analyzed), also in the Bosnjak Kuharic 2021 Cochrane review. According to Cochrane 2021, supported by the US Department of Veterans Affairs and the dronabinol manufacturers (Roxane Laboratories, UNIMED) |
| Bosnjak Kuharic 2021 (Cochrane) Cochrane systematic review and meta-analysis (RCTs only; GRADE; search July 8, 2021) | 4 RCTs, 126 people Dementia (mostly Alzheimer’s; some vascular or mixed) | THC tablet, dronabinol, nabilone | Varies by study Varies by study | 3–14 weeks | Mean difference in behavioral symptoms (NPI) −1.97 (95% CI −3.87 to −0.07; 3 studies, 110 people; low certainty): there may be no clinically important effect. Cognition (sMMSE) +1.1 (0.1 to 2.1; 1 study, 28 people; very low certainty). Authors’ conclusion: it is uncertain whether they are helpful or harmful; any benefit may be too small to be clinically meaningful Side effects: Side effects not reported in a form suitable for meta-analysis; sedation more common with nabilone (OR 2.83; 1.07–7.48); certainty of side effect evidence low to very low | PMID 34532852, doi:10.1002/14651858.CD012820.pub2, PMID 35778327 |
| Tudella 2026 Systematic review and meta-analysis of RCTs (PubMed, Embase, Web of Science, Cochrane; up to November 2025; RoB 2, GRADE, Bayesian analysis) | 9 RCTs, 334 people Alzheimer’s disease or dementia | Various cannabinoids | No effect on the CMAI (SMD −0.58; 95% CI −1.71 to 0.55; I² 84%), NPI-NH total (−0.02; −1.00 to 0.96), NPI-NH agitation (−0.44; −1.45 to 0.57) or MMSE; Bayesian estimates close to zero. Certainty of evidence moderate for behavioral outcomes, low for cognition Side effects: Total side effects similar to placebo; sleepiness more common (RR 2.03; 1.29–3.20) | PMID 42704497, doi:10.1007/s10072-026-09347-z | ||
| Pan 2026 Systematic review and meta-analysis of RCTs (MEDLINE, Embase, PsycINFO, CINAHL, Cochrane, ClinicalTrials.gov; up to August 2025; RoB 2) | 10 RCTs, 328 people Neuropsychiatric symptoms in dementia | Various cannabinoids | No difference in total neuropsychiatric symptoms (SMD −0.18; −0.48 to 0.12); reduction in agitation (SMD −0.52; −1.00 to −0.05; p = 0.03; I² 77%), but significance was lost when studies at high risk of bias were removed (−0.35; −0.79 to 0.10). In subgroups, the effect was larger at doses above 10 mg THC equivalent and in severe dementia. Two of the 8 authors (Ruthirakuhan, Lanctôt) are authors of the nabilone study (Herrmann 2019); no company ties reported; the list of included studies could not be verified because the full text could not be read Side effects: No difference in total side effects; sedation more common (RR 2.09; 1.22–3.57) | PMID 41748849, doi:10.1007/s40263-026-01277-w | ||
| da Silva 2026 Systematic review and meta-analysis (PROSPERO-registered; PubMed, Embase, CENTRAL; up to April 2025; Bayesian analysis and trial sequential analysis) | 6 RCTs and 1 open-label cohort; 221 people Agitation and neuropsychiatric symptoms in Alzheimer’s disease | Various cannabinoids | In randomized comparisons, NPI total (SMD −0.31; −0.47 to −0.15; 4 studies), CMAI short form (−0.40; −0.69 to −0.10; 3 studies) and NPI agitation/aggression (−0.47; −0.69 to −0.25; 3 studies) lower than with placebo; no consistent benefit on the MMSE. According to the authors, interpretation is limited by the small number of studies, different products and scales, short follow-up and the weight being concentrated in a few studies. One of the 14 authors is a consultant to a medical cannabis company (Sensia Labs) Side effects: Main safety signal sleepiness (RR 2.25; 1.43–3.54); falls and fatigue could not be estimated precisely | PMID 42315374, doi:10.1016/j.jagp.2026.05.016 | ||
| Krishnan 2009 (Cochrane) Cochrane systematic review (double- or single-blind, placebo-controlled RCTs; search April 11, 2008) | 1 study (its data could not be analyzed) Dementia | Dronabinol (the only included study) | The data of the only study meeting the criteria (Volicer 1997) had not been reported in a form suitable for analysis; no evidence was found that cannabinoids are effective for disturbed behavior or other symptoms in dementia Side effects: Not reported (in the abstract) | PMID 19370677, doi:10.1002/14651858.CD007204.pub2, The topic was covered by a new Cochrane review in 2021 (Bosnjak Kuharic 2021, CD012820) | ||
| Velayudhan 2021 (safety) Systematic review and meta-analysis of RCTs (mean age ≥50, all indications; 1990–October 2020; GRADE) | 46 RCTs, 6,216 people (3,469 took a cannabinoid; mean age 58.6) Adults over 50 (dementia, Parkinson’s, MS, cancer, etc.) | THC, THC:CBD, CBD | Safety review: THC alone (all-cause side effects IRR 1.42; 1.12–1.78) and THC:CBD (1.58; 1.26–1.98) increased side effects; pure CBD did not (1.02; 0.90–1.16). Serious side effects and deaths did not increase; with THC:CBD, stopping because of side effects increased (RR 1.40; 1.08–1.80), and stopping was linked to the THC dose Side effects: Dry mouth, dizziness and sleepiness more common with THC-containing products | PMID 33780450, doi:10.1371/journal.pmed.1003524, PMID 39602500, 58 RCTs, 6, 611 people, THC dose-dependently increased dizziness, balance and coordination problems, thinking and perception problems, and sleepiness) | ||
| Hatch 2025 (Canadian CCSMH guideline) Clinical practice guideline (Canadian Coalition for Seniors’ Mental Health; review of existing guidelines and systematic reviews; GRADE) | Behavioral and psychological symptoms of dementia (agitation, depression, anxiety, psychosis) | Synthetic cannabinoids (e.g., nabilone) | In severe agitation, synthetic cannabinoids may be used when other drug treatments have not worked (conditional recommendation, low-certainty evidence; C, L). For moderate agitation the first choice is citalopram; for severe agitation aripiprazole, brexpiprazole and risperidone; quetiapine, low-dose typical antipsychotics, synthetic cannabinoids and carbamazepine only if other treatments have not worked. Two of the 17 authors (Herrmann, Lanctôt) are authors of the nabilone study in the table; no author reports ties to a cannabinoid company | PMID 40051590, doi:10.5770/cgj.28.820 | ||
| CDA-AMC 2026 (Canada’s Drug Agency nabilone assessment) Health technology assessment (focused literature search, stakeholder input, 2 clinical experts) | 1 RCT (Herrmann 2019) and 1 guideline Behavioral and psychological symptoms of dementia | Nabilone | Based on a single small study, nabilone may moderately reduce agitation in some Alzheimer’s patients; the evidence is uncertain and it is unclear whether the effect is meaningful for patients; no cost-effectiveness study Side effects: Clinically significant sedation more common | PMID 42485476, CADTH Health Technology Review, NCBI Bookshelf NBK623689 | ||
| Brucki 2021 Position paper of a national academy of neurology (literature assessment by its scientific departments) | Not applicable Neurological diseases; dementia section | Cannabinoids | No evidence supporting clinical use in dementia; use not recommended in Alzheimer’s and related diseases; no evidence that cannabinoids slow the progression of Alzheimer’s, possible benefits are only for symptoms | PMID 34133518, doi:10.1590/0004-282X-ANP-2020-0432, PMC9231442, according to the conflict of interest statement, one of the 49 authors is an executive at a CBD company (HempMeds) | ||
| Alexandri 2024 Controlled, non-randomized open comparison (CBD vs. usual treatment; how allocation was done is not in the abstract) | 20 (10/10) Dementia with severe behavioral symptoms (NPI >30; patients of an Alzheimer’s association in Greece) | 3% CBD drops (THC content not in the abstract) | Not reported (in the abstract) Not reported (in the abstract) | 6 months | NPI improved significantly in all patients taking CBD, with no or limited improvement in the usual treatment group; groups were not formed by chance Side effects: Not reported (in the abstract) | PMID 37153956, doi:10.1080/07317115.2023.2209563 |
| Navarro 2024 Single-center, open-label prospective cohort; no control group | 60 started; 59 followed for at least 3 months, 51 for at least 12 months Alzheimer’s starting after age 65 and neuropsychiatric symptoms not treated with antidepressants, antipsychotics or benzodiazepines; all were taking a cholinesterase inhibitor (Bogotá, Colombia) | CBD-rich compounded (magistral) formula, under the tongue (100 mg CBD + <1.9 mg THC per mL; Khiron; contains THC) | Median 111 mg CBD/day (IQR 79–144.5; at most 196 mg) Started at 0.1 mL every 8–12 hours and increased weekly | Avg. 23.2 months of treatment; primary outcome at week 12 | NPI-Q severity score fell from a median of 24 to 12 at month 3 (p < 0.001); >30% reduction in 94.9%; caregiver distress from 29 to 14; improvement persisted up to 24 months. No comparison group; only those who could pay for the treatment were enrolled; the first author had worked at a clinic owned by the product’s manufacturer Side effects: Sleepiness in 1 patient (product stopped); no other side effects reported | PMID 39474242, doi:10.1159/000541364 |
| Pautex 2022 Prospective observational study (long-term care facility; treatment prescribed by a physician); no control group | 19 Severe dementia, behavior problems and rigidity; older adults taking many medicines (mean age 81.4; 17 women; Geneva) | Standardized cannabis extract, oil or tincture (10 mg THC + 20 mg CBD per mL; THC:CBD 1:2) | At baseline avg. 7.2 mg THC + 14.4 mg CBD/day; at month 13 avg. 12.4 mg THC + 24.8 mg CBD/day (divided into 3 doses a day) Individualized (“start low, go slow”) | Up to 13 months | Marked and sustained improvement in behavior, agitation and rigidity scores and a reduction in other medicines were reported; no comparison group. In 15 patients, CYP1A2 and CYP2C19 enzyme activity was slightly reduced Side effects: 117 side effects, 3 serious events (judged unrelated to treatment); no one stopped because of side effects | PMID 36247984, doi:10.3389/fnagi.2022.957665, PMID 34676334, 10 women, avg. 9.0 mg THC + 18.0 mg CBD/day at month 2, NPI and CMAI fell by about 40%, other psychiatric medicines were reduced or stopped in half), PMID 36967825, the same team’s controlled study: Bianchi 2026 |
| Sawicki 2023 Retrospective chart review (single center, geriatric psychiatry); no control group | 50 (THC recommended); 38 used it Dementia with neuropsychiatric symptoms (Chicago, US) | THC products bought from retail outlets (dispensaries) (the title of the authors’ reply letter mentions edible THC gummies) | Not reported (in the abstract) Not reported (in the abstract) | Records from 2017–2021; follow-up duration not in the abstract | 76% of caregivers bought and gave the product; according to caregivers, symptoms improved in 79% of users (30/38); of the 20 people taking an antipsychotic at baseline, the antipsychotic was reduced or stopped in 12 (60%). No comparison group; assessment based on caregiver reports Side effects: Dizziness, increased agitation and suspiciousness (paranoia); stopped because of side effects in 2 patients | PMID 37728481, doi:10.4088/JCP.23m14791, PMID 38836868 |
| Woodward 2014 Retrospective systematic chart review (inpatient unit); no control group | 40 Inpatients with severe dementia given dronabinol for behavior or appetite problems (geriatric neuropsychiatry unit, US) | Dronabinol (added to existing treatment) | Not reported (in the abstract) Not reported (in the abstract) | Up to 7 days after starting dronabinol | Improvement in all domains of the Pittsburgh Agitation Scale, in the clinical global impression, in sleep duration and in the share of meals eaten; no comparison group Side effects: 26 side effects, none led to stopping treatment | PMID 23597932, doi:10.1016/j.jagp.2012.11.022 |
| Palmieri 2023 Retrospective observational study; no control group | 30 Mild, moderate or severe Alzheimer’s, aged 65–90 (Italy, second-opinion clinic) | Bedrocan cannabis extract diluted in olive oil (22% THC, 0.5% CBD) | Twice a day; amount not in the abstract Not reported (in the abstract) | 12 weeks | On the NPI-Q filled in by relatives, reductions in agitation, apathy, irritability, sleep and eating disorders, and on the CMAI, in aggressive behavior (p < 0.0001); the MMSE was reported to have improved in 45% of patients. The authors note the lack of a control group and the fact that the scales were filled in by relatives as limitations Side effects: Not reported (in the abstract) | PMID 36655645, doi:10.7417/CT.2023.5009 |
| Myran 2025 Population-based, retrospective, matched cohort (Ontario health records, 2008–2021) | 6,086,794 people; 16,275 with an emergency visit or hospital admission due to cannabis Ontario residents aged 45–105 without dementia at baseline | Cannabis use (at a level requiring emergency care) | Not applicable | Follow-up of up to 5 years (to 2022) | Dementia diagnosis within 5 years: 5.0% in those receiving emergency care due to cannabis, 3.6% in those receiving emergency care for other reasons, 1.3% in the general population. Adjusted hazard ratio 1.23 (1.09–1.39) vs. emergency care for other reasons, 1.72 (1.38–2.15) vs. the general population; 0.69 (0.62–0.76) vs. alcohol-related emergency care. Observational: does not show cause and effect; confounding factors cannot be ruled out Side effects: Not applicable | PMID 40227745, doi:10.1001/jamaneurol.2025.0530 |
| Ishrat 2026 Observational cohort analyses (UK Biobank, US Million Veteran Program) and bidirectional Mendelian randomization | In UK Biobank, up to 18,975 people who had used cannabis and 60,598 who had not; in the MVP, 12,222 people diagnosed with cannabis use disorder Older adults | Cannabis use (lifetime) and cannabis use disorder | Not applicable | Longitudinal follow-up of cognition and dementia (duration not in the abstract) | People who had used cannabis did slightly better at baseline on numeric memory and reasoning tests; no difference in longitudinal cognitive change. Cannabis use disorder not significantly linked to dementia risk (HR 1.11; 0.97–1.26). Mendelian randomization showed no causal link between cannabis use and cognition or dementia Side effects: Not applicable | PMID 41741130, doi:10.1136/bmjment-2025-302290 |
Reported side effects
- Sleepiness about twice as common with cannabinoids: RR 2.03 (Tudella 2026), 2.09 (Pan 2026), 2.25 (da Silva 2026); with nabilone 45% vs. 16% (Herrmann 2019)
- With the CBD-rich, THC-containing oil: memory loss 32% vs. 10%, hallucinations 22% vs. 5%, sleepiness 49% vs. 30%; all 8 patients who dropped out were in the treatment group (Hermush 2022)
- With pure CBD (up to 600 mg a day): dizziness 63% vs. 0% (8 and 7 people); falls 25% vs. 43%; raised liver enzymes (Velayudhan 2024, CANBiS-AD)
- With dronabinol: serious events 2 vs. 0 (delirium, broken arm bone), diarrhea 5 vs. 0, falls 3 vs. 5 (NCT02792257 registry results; Rosenberg 2026)
- With the low-dose THC tablet, body sway with the eyes closed increased; no falls (Ahmed 2015; van den Elsen 2017)
- 46 RCTs in people over 50: THC alone (IRR 1.42) and THC:CBD (IRR 1.58) increased all side effects, while pure CBD did not (IRR 1.02) (Velayudhan 2021)
Monitoring done in studies
- If a cannabinoid is used in a person with dementia, sleepiness, falls, confusion, blood pressure and appetite should be monitored with the doctor
- With high-dose or long-term CBD use, liver tests are the doctor’s decision (raised liver enzymes were reported with up to 600 mg of pure CBD a day; Velayudhan 2024)
- If other medicines are taken, drug interactions should be assessed with the pharmacist and doctor
Who should be careful
- Sleepiness and falls: in three meta-analyses pooling dementia studies, sleepiness was about twice as common with cannabinoids (Tudella 2026, Pan 2026, da Silva 2026); in the dronabinol study, a broken arm bone and an episode of sudden confusion (delirium) were recorded as serious events in the treatment group, while the number of falls was similar in the two groups (3 vs. 5 people; NCT02792257 registry results)
- THC, memory and hallucinations: with the CBD-rich oil that reached an average of 22 mg of THC a day, memory loss (32% vs. 10%) and hallucinations (22% vs. 5%) were numerically more common; the difference was right at the edge of statistical significance (Hermush 2022)
- Labeling: a 94-year-old woman whose family gave her a product sold as “pure CBD” developed THC-related confusion and hallucinations (case report; Moussa 2023, PMID 37223146)
- High-dose pure CBD: with up to 600 mg of CBD a day, dizziness (63%) and raised liver enzymes were seen (Velayudhan 2024, CANBiS-AD)
- Blood pressure: blood pressure changes beyond the protocol limit were recorded more often during the period on the oil containing THC and CBD than during the placebo period (71 vs. 28 events; about half were drops and half were rises in blood pressure, average values did not change in either period, and no statistical test was done; Bianchi 2026)
- Psychiatric medicines: most patients in the studies continued their antidepressants, antipsychotics or sedatives; it has not been shown that cannabinoids can replace these medicines. In one controlled study, the use of sedatives given as needed went down (Bianchi 2026)
Drug interactions
- In 15 patients with severe dementia who used an oil containing THC and CBD for months, the activity of the enzymes CYP1A2 and CYP2C19 was slightly reduced; these enzymes help break down certain medicines in the body (Pautex 2022)
- In a controlled study with an oil containing THC and CBD for 8 weeks in patients with severe dementia taking many medicines, no problem due to drug interactions was reported (Bianchi 2026)
- Dementia studies did not systematically examine interactions with antidepressants, antipsychotics and sedatives; details are in the site’s Drug interactions section (antidepressants, antipsychotics and lithium, anxiety and sleep medicines)
Abbreviations: RCT, randomized controlled trial (a study in which participants are assigned to groups by chance); placebo, an inactive substance; SMD, OR and RR, effect measures used in analyses that pool studies; CYP, liver enzymes that break down medicines.
If you are considering CBD
- Do not stop or change your loved one’s dementia, sedative, sleep or antipsychotic medicines on your own; there is no evidence that CBD or cannabis products can replace them.
- If you are considering a product, talk to your loved one’s doctor first; review all the medicines they take together with the doctor and pharmacist.
- Guidelines recommend looking together at the physical, mental and environmental factors that contribute to agitation, and trying non-drug approaches first; you can talk about these with the doctor too.
- If a product has been started, note any sleepiness, drowsiness, falls, increased confusion and changes in appetite, and share them with the doctor.
- Check the product label and certificate of analysis (COA); even products sold as “pure CBD” can contain THC.
Choose your age group, medicines and situation to prepare a short summary you can take to your doctor. Nothing you enter is sent anywhere.
Bottom line
Studies with cannabinoids for agitation and behavioral symptoms in dementia are small, short and contradict each other; with pure CBD there are only two small studies. The most consistent finding is increased sleepiness. For now these products do not count as a proven treatment; talking with your loved one’s doctor before trying anything is the safest path.
Frequently asked questions
Does CBD oil work for dementia or Alzheimer’s?
Current studies do not show this clearly. There are only two small studies with pure CBD: in 30 people with dementia caused by blood vessel problems, behavioral symptoms went down slightly more than with an inactive pill (placebo), but the difference was right at the edge of statistical significance; the 15-person pilot study was not designed to test whether it works. Of the studies with products rich in CBD but also containing THC, one had a positive result and several had negative results. It has not been shown that CBD improves memory and thinking skills; the positive mental test result in a 28-person study with an extract containing very low doses of THC and CBD has not yet been confirmed by another study. For pure CBD the strength of evidence is very weak; for THC-containing products the results are mixed and the evidence is weak.
Does CBD reduce agitation and aggression in Alzheimer’s?
It is not certain. There are only two small studies with pure CBD (30 and 15 people); in both, behavioral scores improved numerically more with CBD, but in one the difference was borderline, and the other was not designed to test whether it works. In some of the controlled studies of THC-containing products for agitation (the CBD-rich oil that also contains THC, with a borderline difference on the main measure; nabilone; one of the two scales with dronabinol), agitation went down more than with placebo, while in others (low-dose THC tablets, an oil and a spray containing THC and CBD) there was no difference. Of three 2026 meta-analyses pooling these studies, one found no effect and two found a small to moderate effect, and in one of these the effect was no longer significant when studies at high risk of bias were removed. The strength of evidence is very weak for pure CBD and weak for THC-containing products.
Does CBD prevent or slow Alzheimer’s?
No human study shows this; the idea comes from animal and laboratory experiments. A 236-person study testing CBD in people with mild cognitive impairment is under way. In large observational studies, having used cannabis was not significantly linked to cognitive decline, and cannabis use disorder was not significantly linked to dementia risk; on the other hand, middle-aged and older adults treated in emergency care because of cannabis were diagnosed with dementia more often in the following years. These studies do not show cause and effect. The Brazilian Academy of Neurology also writes that there is no evidence that cannabinoids slow the progression of Alzheimer’s.
Are CBD or cannabis products safe for an older loved one?
The most common side effect in the studies was sleepiness; it was about twice as common in people taking cannabinoids. With THC-containing products, drowsiness, confusion, memory loss and hallucinations were reported; with high-dose pure CBD, dizziness and raised liver enzymes were seen. These side effects matter because the risk of falls is already high in older adults with dementia. Even products sold as “pure CBD” can contain THC. So if a product is being considered, talking first with your loved one’s doctor and pharmacist, and if it is being used, watching for sleepiness, falls and confusion, is the safest path.
Can CBD be used instead of antipsychotics or sedatives?
No study shows this. In the controlled studies, most patients continued these medicines and cannabinoids were added to them. In a study in Geneva with nursing home residents with severe dementia, the use of sedatives given as needed went down during the period on the oil containing THC and CBD; in a US chart review without a control group, the antipsychotic could be reduced in more than half of those taking one. Canada’s 2025 guideline gives synthetic cannabinoids a place only in severe agitation that has not responded to other medicines, conditionally and based on low-quality evidence. On this question the strength of evidence is very weak. Do not stop or change medicines on your own.
Does CBD help with sleep problems or loss of appetite in dementia?
There is no large study designed specifically for these questions. In a 60-person study with a CBD-rich oil that also contains THC, sleep problems went down more in the treatment group, but complaints of sleepiness were also more common in the same group; in a 75-person study with dronabinol, there was no difference in sleep. For appetite and weight, the most direct data come from a small 1997 study with 15 Alzheimer’s patients who refused to eat: weight went up more during the dronabinol period. There is no controlled study of pure CBD on sleep or appetite in dementia. The strength of evidence is very weak.
Can CBD oil for dementia be obtained with a prescription in Turkey?
In Turkey, products derived from cannabis are regulated by the Kenevirden Elde Edilen Ürünlere Dair Yönetmelik (Regulation on Products Derived from Hemp) dated January 31, 2026. These products, including CBD, can be sold only in pharmacies. Those classed as medicines and health products are dispensed with a prescription; the regulation does not state a prescription requirement for supplement products and cosmetics with a THC content of at most 0.3%. As of October 2026, there is no approved (licensed) CBD medicine in Turkey. Dronabinol (synthetic THC) and nabilone (a synthetic THC-like medicine), which were tested in dementia studies, are not licensed in Turkey either: there is no cannabinoid-containing medicine on the list of licensed medicines of the Turkish Medicines and Medical Devices Agency (TİTCK) (October 16, 2026), and dronabinol and nabilone are not on the agency's list of active substances that can be brought in from abroad (September 28, 2026) either. If you are thinking of using it, talk to your loved one’s doctor first.
References
Show 29 references
- Rosenberg 2026 (THC-AD). Phase 2, multicenter (5 centers), double-blind, placebo-controlled parallel RCT. PMID 41350162, doi:10.1016/j.jagp.2025.10.011, NCT02792257, May 9, 2025, 84 participants in the registry, 75 randomized in the publication), PMID 34629131, PMID 41350163
- Hermush 2022. Single-center, double-blind, placebo-controlled parallel RCT (2:1). PMID 36148467, doi:10.3389/fmed.2022.951889, NCT03328676
- van den Elsen 2015a. Phase 2, multicenter, double-blind, placebo-controlled parallel RCT. PMID 25972490, doi:10.1212/WNL.0000000000001675, NCT01608217, patients with pain or suspected pain (29 people) were also given paracetamol (acetaminophen), the drug was supplied by the manufacturer (Echo Pharmaceuticals), funding from the EU Regional Development Fund and the province of Gelderland
- Herrmann 2019. Phase 2/3, double-blind, placebo-controlled crossover RCT (2 × 6 weeks, with a 1-week washout in between). PMID 31182351, doi:10.1016/j.jagp.2019.05.002, NCT02351882, PMID 31338476, PMID 31322567, 24S-hydroxycholesterol), PMID 31547752, oxidative stress and cytokines), PMID 41530008, in a post hoc analysis, response was larger in those with marked pain, appetite and eating problems, or apathy), PMID 41220257, predicting response with blood markers, exploratory), PMID 31196620
- Pessoa 2026. Double-blind, placebo-controlled parallel RCT. PMID 41277041, doi:10.1177/02698811251390974, ReBEC RBR-686kmpz
- Cury 2025. Phase 2, double-blind, placebo-controlled parallel RCT. PMID 41160460, doi:10.1177/13872877251389608, ReBEC U1111-1258-2058, PMID 42216668
- Albertyn 2025 (STAND). Double-blind, placebo-controlled feasibility RCT (nursing homes in England). PMID 40479610, doi:10.1093/ageing/afaf149, ISRCTN 7163562 (as written in the publication), drug and placebo from Jazz Pharmaceuticals (formerly GW Pharmaceuticals), funding from Alzheimer's Research UK, one of the 12 authors has served on advisory boards of cannabis companies (EmpowerPharm/SanteCannabis, NW PharmaTech) or received research support through their institution
- Bianchi 2026 (MedCanDem). Multicenter, double-blind, placebo-controlled crossover RCT (2 × 8 weeks, 1-week washout). PMID 42613673, doi:10.1093/ageing/afag239, NCT05432206, PMID 37293434, product and placebo provided free by the manufacturer, two authors are former board members of the Swiss Society for Cannabis in Medicine, the same team’s uncontrolled follow-up: Pautex 2022
- van den Elsen 2015b. Phase 2, multicenter, double-blind, placebo-controlled, repeated crossover RCT (6 blocks × 2 weeks: in each block 3 days of THC and 3 days of placebo, with 4-day washouts in between). PMID 26560511, doi:10.1016/j.jagp.2015.07.011, NCT01302340, PMID 25752889, PMID 27624148, a secondary analysis of the same trial according to the Bahji 2022 review, could not be verified from the full text)
- Timler 2023. Double-blind, placebo-controlled crossover RCT (2 × 6 weeks of treatment, with 2-week washouts). PMID 37321847, doi:10.1111/ajag.13224, ANZCTR ACTRN12619000474156, PMID 32059690, funded by the product’s manufacturer, MGC Pharmaceuticals)
- Velayudhan 2024 (CANBiS-AD). Phase 2a, single-center, double-blind, placebo-controlled parallel feasibility RCT (published as a letter to the editor). PMID 38584388, doi:10.1017/S1041610224000516, EudraCT 2019-002106-52, study drug provided free by Beckley Canopy Therapeutics/Canopy Growth (investigator-initiated study), one of the 7 authors has served on advisory boards of cannabis companies (EmpowerPharm/SanteCannabis) or received speaker fees, funding from King's Health Partners, Psychiatry Research Trust and Rosetrees Trust
- Volicer 1997. Placebo-controlled crossover study (2 × 6 weeks; controlled clinical trial in PubMed; randomized and double-blind according to reviews). PMID 9309469, The only study included in the Krishnan 2009 Cochrane review (its data could not be analyzed), also in the Bosnjak Kuharic 2021 Cochrane review. According to Cochrane 2021, supported by the US Department of Veterans Affairs and the dronabinol manufacturers (Roxane Laboratories, UNIMED)
- Bosnjak Kuharic 2021 (Cochrane). Cochrane systematic review and meta-analysis (RCTs only; GRADE; search July 8, 2021). PMID 34532852, doi:10.1002/14651858.CD012820.pub2, PMID 35778327
- Tudella 2026. Systematic review and meta-analysis of RCTs (PubMed, Embase, Web of Science, Cochrane; up to November 2025; RoB 2, GRADE, Bayesian analysis). PMID 42704497, doi:10.1007/s10072-026-09347-z
- Pan 2026. Systematic review and meta-analysis of RCTs (MEDLINE, Embase, PsycINFO, CINAHL, Cochrane, ClinicalTrials.gov; up to August 2025; RoB 2). PMID 41748849, doi:10.1007/s40263-026-01277-w
- da Silva 2026. Systematic review and meta-analysis (PROSPERO-registered; PubMed, Embase, CENTRAL; up to April 2025; Bayesian analysis and trial sequential analysis). PMID 42315374, doi:10.1016/j.jagp.2026.05.016
- Krishnan 2009 (Cochrane). Cochrane systematic review (double- or single-blind, placebo-controlled RCTs; search April 11, 2008). PMID 19370677, doi:10.1002/14651858.CD007204.pub2, The topic was covered by a new Cochrane review in 2021 (Bosnjak Kuharic 2021, CD012820)
- Velayudhan 2021 (safety). Systematic review and meta-analysis of RCTs (mean age ≥50, all indications; 1990–October 2020; GRADE). PMID 33780450, doi:10.1371/journal.pmed.1003524, PMID 39602500, 58 RCTs, 6, 611 people, THC dose-dependently increased dizziness, balance and coordination problems, thinking and perception problems, and sleepiness)
- Hatch 2025 (Canadian CCSMH guideline). Clinical practice guideline (Canadian Coalition for Seniors’ Mental Health; review of existing guidelines and systematic reviews; GRADE). PMID 40051590, doi:10.5770/cgj.28.820
- CDA-AMC 2026 (Canada’s Drug Agency nabilone assessment). Health technology assessment (focused literature search, stakeholder input, 2 clinical experts). PMID 42485476, CADTH Health Technology Review, NCBI Bookshelf NBK623689
- Brucki 2021. Position paper of a national academy of neurology (literature assessment by its scientific departments). PMID 34133518, doi:10.1590/0004-282X-ANP-2020-0432, PMC9231442, according to the conflict of interest statement, one of the 49 authors is an executive at a CBD company (HempMeds)
- Alexandri 2024. Controlled, non-randomized open comparison (CBD vs. usual treatment; how allocation was done is not in the abstract). PMID 37153956, doi:10.1080/07317115.2023.2209563
- Navarro 2024. Single-center, open-label prospective cohort; no control group. PMID 39474242, doi:10.1159/000541364
- Pautex 2022. Prospective observational study (long-term care facility; treatment prescribed by a physician); no control group. PMID 36247984, doi:10.3389/fnagi.2022.957665, PMID 34676334, 10 women, avg. 9.0 mg THC + 18.0 mg CBD/day at month 2, NPI and CMAI fell by about 40%, other psychiatric medicines were reduced or stopped in half), PMID 36967825, the same team’s controlled study: Bianchi 2026
- Sawicki 2023. Retrospective chart review (single center, geriatric psychiatry); no control group. PMID 37728481, doi:10.4088/JCP.23m14791, PMID 38836868
- Woodward 2014. Retrospective systematic chart review (inpatient unit); no control group. PMID 23597932, doi:10.1016/j.jagp.2012.11.022
- Palmieri 2023. Retrospective observational study; no control group. PMID 36655645, doi:10.7417/CT.2023.5009
- Myran 2025. Population-based, retrospective, matched cohort (Ontario health records, 2008–2021). PMID 40227745, doi:10.1001/jamaneurol.2025.0530
- Ishrat 2026. Observational cohort analyses (UK Biobank, US Million Veteran Program) and bidirectional Mendelian randomization. PMID 41741130, doi:10.1136/bmjment-2025-302290
Points that could not be verified from the full text
- Full text could not be read (October 28, 2026): van den Elsen 2015b and 2017, Herrmann 2019, Rosenberg 2026 (full text under publisher embargo until November 10, 2026; only the abstract and ClinicalTrials.gov results), Timler 2023 (abstract and protocol), Cury 2025 (abstract and a commentary), Pessoa 2026, Volicer 1997, Alexandri 2024, Woodward 2014, Sawicki 2023, Palmieri 2023, the 2009 Cochrane review and the three meta-analyses from 2026 (abstract only). Details of Volicer 1997 were taken from the tables of systematic reviews; its dose is given differently across reviews, and according to the study table of the 2021 Cochrane review it was 2.5 mg in the morning and at noon (5 mg a day).
- Controlled studies completed but with no results released as of October 2026: NCT06570928 (Brazil; 0.2 mL a day of an oil containing 50 mg of CBD and 5 mg of THC per mL, that is, 10 mg of CBD and 1 mg of THC; main measure a mental test; 72 people with Alzheimer’s; finished in March 2025), NCT05644262 (US; oil containing THC and CBD; agitation in 120 people with dementia in the last stage of life who were eligible for hospice care; finished in April 2026), NCT04749563 (IGC-AD1 containing low-dose THC; 12-person safety study; 2021). Stopped early: NCT05742698 (nabilone in frontotemporal dementia; only 3 people randomized, terminated), NCT06217146 (cannabis oil; only the open-label first part was completed). Status unknown: NCT05239390 (SCI-110 containing dronabinol), NCT05022186 (CBD in mild cognitive impairment), NCT06514066 (CBD in vascular and mixed dementia).
- Ongoing or not yet started studies (October 28, 2026): NAB-IT NCT04516057 (nabilone for agitation in Alzheimer’s, phase 3, estimated April 2027), NCT05543681 (IGC-AD1 containing low-dose THC, 164 people, estimated November 2026), NCT05822362 (CBD in mild cognitive impairment, 236 people, estimated 2029), CALM-IT NCT06014424 (CBD for agitation, 40 people, estimated December 2026), NCT04436081 (THC-free CBD oil, 40 people; registry entry not updated since 2023), NCT04075435 (CBD-rich solution with low THC, open-label, 12 people), NCT07422311 (dronabinol, 140 people, not yet started), NCT07091747 (open-label extension of NCT06570928, 61 people). Those without a registration number found or listed in other registries: pure CBD in early-stage dementia in Australia (ACTRN12621001364864; protocol Bartschi 2023) and a mouth spray containing THC and CBD for pain and agitation in severe dementia in Italy (protocol Scuteri 2022).
- The dementia chapter of the 2017 US National Academies report and the full text of Canada’s Drug Agency 2026 report could not be read because NCBI Bookshelf required bot verification; only the summary of the latter was used.
- The status of dronabinol and nabilone in Turkey was read from two lists of TİTCK (Turkish Medicines and Medical Devices Agency): the List of Licensed Medicinal Products for Human Use (October 16, 2026) and the List of Active Substances from Abroad (September 28, 2026), both downloaded on October 29, 2026. Whether a drug that is not on the list could be brought in some other way was not investigated in this review; because no current version of the Turkish Pharmacists' Association's list of medicines from abroad could be found, that list was not searched.
Research history
Here we record what changed on this page, when and why, with dates. The newest change is at the top.
The last sentence of the page's short summary was rewritten in a calmer, more guiding tone: instead of "Talk to your doctor before you start.", it now says whom you can talk with, and about what, if you are considering CBD. The conclusion and the strength of evidence did not change.
It was added to the question about Turkey that dronabinol and nabilone, which were tested in dementia studies, are not licensed in Turkey: there is no cannabinoid-containing medicine on the Turkish Medicines and Medical Devices Agency's list of licensed medicines (October 16, 2026), and neither drug appears on the agency's list of active substances from abroad (September 28, 2026). The conclusion and the strength of evidence did not change.
Page published. In this first assessment, 29 studies were reviewed.