THC and CBD for Dementia Agitation: What the LiBBY Trial Actually Found

A hopeful cannabis result becomes less useful the moment it is made to claim too much.

The LiBBY trial offers genuinely encouraging evidence that a carefully formulated combination of THC and CBD may reduce agitation in people living with advanced dementia near the end of life. It was randomised, double-blind and placebo-controlled. It included 120 participants, met its primary endpoint and reported a substantial difference between the treatment and placebo groups within two weeks.

That is a serious finding in a population too often excluded from clinical research. It is not evidence that cannabis treats dementia, restores memory or slows neurodegeneration. Nor is it a reason for families to experiment with CBD oils or cannabis products bought online.

The value of LiBBY lies in the narrow question it asked: can one precise, medically supervised oral formulation ease a distressing symptom when options are limited?

What the trial tested

LiBBY—short for Life's End Benefits of Cannabidiol and Tetrahydrocannabinol—was a 12-week phase-two study conducted across ten sites in the United States. The results were presented at the Alzheimer's Association International Conference in London on 14 July 2026.

Researchers enrolled 120 people with Alzheimer's disease or another condition causing dementia. Participants were experiencing clinically significant agitation and were either receiving hospice care or considered eligible for it. Their average age was about 80, and three-quarters lived in community settings rather than residential facilities.

Sixty-one participants were assigned to the investigational medicine and 59 to placebo. Neither participants, caregivers nor clinicians knew which treatment had been allocated during the controlled phase.

The medicine, known as T2:C100, was not a general cannabis oil. Each millilitre contained 2 mg of purified THC and 100 mg of CBD in a digestible oil. Participants began with 1 ml twice daily and increased to 2 ml twice daily, giving a divided daily dose of 8 mg THC and 400 mg CBD at the full dose.

The trial team is explicit that the formulation is not publicly available and that its findings do not apply to other cannabis products.

What the topline results showed

The primary measure was change at two weeks on the Cohen-Mansfield Agitation Inventory, a structured scale covering the frequency of behaviours associated with agitation.

At week two, the treated group recorded a 6.27-point greater reduction in agitation than the placebo group. The reported 95% confidence interval ranged from 2.87 to 9.66 points in favour of treatment, and the difference was statistically significant.

The benefit was sustained across the 12-week controlled phase. The trial presentation reported an 8.23-point difference between groups on its week-12 analysis.

Clinician ratings pointed in the same direction. At two weeks, 83.9% of people receiving T2:C100 were rated as improved, compared with 30.5% on placebo. At 12 weeks, the figures were 87.2% and 23.6% respectively.

These are notable results. Agitation in advanced dementia can include pacing, repetitive movements, persistent calling out, emotional distress and physical or verbal aggression. It can also be a sign of pain, fear, delirium, discomfort or another unmet need that a person can no longer communicate easily.

Reducing that distress may improve the experience of the person, their family and those providing care. It should not be confused with changing the underlying dementia.

The safety findings need a full reading

Overall adverse-event rates were similar: 46.7% in the treatment group and 42.4% with placebo. That headline is reassuring, but it is not the whole safety picture.

Serious adverse events occurred in 23.3% of treated participants and 11.9% of those receiving placebo. Eight people in the treatment arm and three in the placebo arm died during the 12-week phase. The investigators reported that none of the serious events or deaths was related to the study medicine and said there was no pattern in the causes of death.

This was a medically fragile, end-of-life population, so deaths and serious illness were expected. Even so, the numerical imbalance should not be brushed aside. A complete peer-reviewed paper will need to show how events were assessed, when they occurred, what competing clinical factors were present and how robust the safety conclusions remain under different analyses.

Treatment discontinuation was also higher with T2:C100: 26.2% compared with 10.2% on placebo by week 12. Most recorded discontinuations in both groups were attributed to death or being unable or unwilling to continue rather than a listed adverse event, but the difference matters when judging how feasible treatment is in ordinary care.

The trial's 12-week open-label extension reportedly found continued improvement among those remaining on treatment and improvement among participants who switched from placebo. Because everyone knew they were receiving the active medicine and there was no continuing placebo comparison, that phase is supportive evidence rather than a replacement for the controlled result.

What LiBBY does not establish

The study did not test whether THC and CBD prevent dementia, slow its progression or improve memory. It examined agitation in people with severe, late-stage illness.

It also did not test cannabis flower, high-street CBD, homemade preparations or the range of products available through medical-cannabis clinics. Dose, ratio, purity, delivery method and clinical monitoring are all part of the intervention.

This distinction is especially important because the findings are currently topline conference results. The investigators have published detailed presentation slides, but a complete peer-reviewed report is still needed. Conference data can be important and well conducted; it simply has not yet received the scrutiny, methodological detail and permanence of a journal publication.

The comparison with placebo also does not mean medication should be the first response to every distressed behaviour. Agitation is a description, not a diagnosis. Pain, infection, constipation, delirium, overstimulation, loneliness, an unfamiliar setting or unmet personal needs may all require a different response.

What this means in a UK context

Current NICE guidance says clinicians should first conduct a structured assessment for clinical and environmental causes of distress. Psychosocial and environmental interventions should be offered as initial and continuing management.

Antipsychotics should be reserved for people who risk harming themselves or others, or whose agitation, hallucinations or delusions are causing severe distress. When used, NICE recommends the lowest effective dose for the shortest possible time, with reassessment at least every six weeks.

LiBBY does not rewrite that pathway. It gives researchers a credible reason to test T2:C100 or closely controlled formulations in a larger and broader population. Any move towards UK practice would require a full evidence review, medicine-quality controls, regulatory consideration and condition-specific prescribing guidance.

Most cannabis-based products for medicinal use remain unlicensed in the UK. A new indication cannot be inferred from one US phase-two trial, however positive the topline result appears.

For families, the practical message is not to obtain THC or CBD independently. Older people with advanced dementia are particularly vulnerable to sedation, dizziness, confusion, falls and medicine interactions. Any distressing change in behaviour needs clinical assessment, especially when it appears suddenly.

Why this research still matters

People with advanced dementia are commonly excluded from trials because consent, frailty, communication and follow-up are difficult. Those difficulties do not make their symptoms less deserving of evidence.

LiBBY showed that a controlled trial can recruit and retain people near the end of life while involving caregivers and delivering study visits where participants live. That may be one of its most lasting contributions.

It also put comfort at the centre of the research question. In late-stage dementia, a worthwhile treatment may not alter a scan or cognitive test. Its value may be measured in fewer frightened movements, less persistent distress, a calmer interaction or an easier period of care.

Those outcomes are clinically meaningful. They also require careful interpretation because many are observed and rated by caregivers or clinicians rather than reported directly by participants.

What should happen next

The first priority is full peer-reviewed publication of the controlled and extension data. Researchers should make the protocol, statistical plan, adverse-event adjudication and missing-data handling easy to examine.

A larger trial should test whether the effect is replicated across different dementia diagnoses, care settings and levels of frailty. It should identify which behaviours respond, whether benefits reduce caregiver burden, how alertness and falls are affected, and whether a lower dose could retain benefit with fewer risks.

Comparison with current practice will eventually matter too. Placebo answers whether the formulation has an effect. Patients and health systems also need to know how it compares with carefully delivered non-drug care and with medicines already used for severe agitation.

Our wider guide to cannabinoid research in 2026 explains why replication, formulation and clinically meaningful outcomes matter as much as a positive headline.

The editorial view

LiBBY deserves attention because it studied a neglected population with a rigorous design and asked a humane question.

Its findings are strong enough to justify a larger trial. They are not strong enough to turn an experimental medicine into general advice about cannabis and dementia.

The responsible response is neither dismissal nor celebration without limits. It is to protect the precision of the result: one purified THC/CBD formulation, one defined population, one distressing symptom and one promising controlled trial that now needs independent scrutiny and replication.

For families confronting advanced dementia, hope should come with clarity. This research may point towards a better option in future. It is not an option they can safely recreate at home today.

Key takeaways

• LiBBY was a 12-week, randomised, double-blind, placebo-controlled phase-two trial involving 120 hospice-eligible people with advanced dementia and clinically significant agitation.
• A precise oral formulation containing purified THC and CBD produced a greater reduction in agitation than placebo at two weeks, with benefits sustained through week 12.
• Clinician-rated improvement was reported in 87.2% of treated participants and 23.6% of the placebo group at week 12.
• Overall adverse-event rates were similar, but serious events, deaths and discontinuations were numerically higher in the treatment group and require close examination in the full publication.
• The trial did not test whether cannabinoids treat or slow dementia, and its results do not apply to retail CBD, cannabis flower or other formulations.
• UK guidance still prioritises assessment of underlying causes and psychosocial or environmental support before medication for agitation.

Sources

1. LiBBY double-blind phase topline results — Alzheimer’s Clinical Trial Consortium
2. LiBBY study results and formulation warning — LiBBY Study
3. THC/CBD combination and late-stage dementia agitation — Alzheimer’s Association International Conference
4. LiBBY trial report — Georgetown University School of Medicine
5. LiBBY trial registration NCT05644262 — ClinicalTrials.gov
6. Dementia: assessment, management and support — NICE guideline NG97