Neurology · Medical Literature
Cannabinoids Treatment for Agitation in Alzheimer's Disease: A Systematic Review and Meta-Analysis With Bayesian and Sequential Trial Analyses.
Abstract
Agitation and related neuropsychiatric symptoms are common in Alzheimer's disease (AD) and contribute to caregiver burden, functional decline, and institutionalization. Cannabinoid-based therapies have been investigated as potential symptomatic interventions, but evidence remains limited by small trials, heterogeneous formulations, and variable outcome reporting. We aimed to evaluate the efficacy, cognitive outcomes, and safety of cannabinoid-based therapies in AD. This systematic review and meta-analysis were prospectively registered in PROSPERO and conducted following PRISMA 2020 guidelines. PubMed, Embase, and the Cochrane Central Register of Controlled Trials were searched through April 2025 for randomized placebo-controlled trials evaluating cannabinoid-based therapies for agitation or neuropsychiatric symptoms in AD. One nonrandomized open-label study was retained as supplementary evidence for sensitivity analyses. Primary efficacy analyses were restricted to randomized between-group contrasts. Outcomes included Neuropsychiatric Inventory (NPI) total score, Cohen-Mansfield Agitation Inventory-Short Form (CMAI-SF), NPI agitation/aggression, Mini-Mental State Examination (MMSE), and safety outcomes. Random-effects meta-analyses were complemented by Bayesian models and Trial Sequential Analysis. Seven studies met the inclusion criteria, including six randomized trials and one open-label prospective cohort, with 221 participants enrolled. Cannabinoid-based therapies showed lower neuropsychiatric symptom and agitation scores than placebo in randomized between-group analyses: NPI total score (standardized mean difference [SMD], -0.31; 95% confidence intervals [CI], -0.47 to -0.15; k = 4), CMAI-SF (SMD, -0.40; 95% CI, -0.69 to -0.10; k = 3), and NPI agitation/aggression (SMD, -0.47; 95% CI, -0.69 to -0.25; k = 3). Bayesian posterior probabilities for lower symptom scores exceeded 95%. MMSE findings did not support a consistent cognitive benefit. Somnolence was the principal safety signal (risk ratios, 2.25; 95% CI, 1.43-3.54), with Trial Sequential Analysis suggesting sufficient accrued information for this outcome. Falls and fatigue were imprecisely estimated. Cannabinoid-based therapies showed lower agitation and neuropsychiatric symptom scores than placebo in AD, with somnolence as the main safety concern. Interpretation remains limited by few trials, heterogeneous formulations and outcome instruments, short follow-up, and concentrated statistical weight. Larger randomized trials with formulation-specific protocols, longer follow-up, active comparators, and systematic safety monitoring are needed.
Clinical relevance
In adults with moderate-to-severe Alzheimer's disease, do cannabinoid-based therapies reduce agitation and neuropsychiatric symptoms more effectively than placebo, and what is their cognitive and safety profile? __SNAPSHOT__ {"studyDesign":"Systematic review and meta-analysis with supplementary Bayesian and Sequential Trial Analyses (Trial Sequential Analysis); prospectively registered in PROSPERO; conducted per PRISMA 2020 guidelines. Primary efficacy analyses restricted to randomized between-group contrasts (k=6 RCTs); one non-randomized open-label study retained for sensitivity analyses only.","setting":"Multinational; 6 countries; variety of care facilities (specific institution types not reported). Literature search through April 2025.","sampleSize":"k=7 studies total (6 RCTs + 1 open-label cohort); 221 participants; mean age range 72.7–87.0 years","population":"Adults with moderate-to-severe Alzheimer's disease with clinically documented agitation or validated neuropsychiatric symptoms","intervention":"Cannabinoid-based therapies including: dronabinol (2.5–10 mg/day), nabiximols (THC 10.8 mg + CBD 10 mg/day), nabilone (1–2 mg/day), THC medical cannabis oil (2.5–7.5 mg/day), and THC-CBD extract (0.35 mg THC + 0.25 mg CBD/day)","comparator":"Placebo","primaryOutcome":"Global neuropsychiatric burden (NPI total score), Cohen-Mansfield Agitation Inventory-Short Form (CMAI-SF), NPI agitation/aggression domain score, MMSE (cognitive safety), and safety outcomes","followUp":"Trial durations ranged from 2 to 26 weeks","fundingSource":"Not reported"}
Key findings
In randomized between-group analyses, cannabinoid-based therapies were associated with statistically significantly lower scores across all three neuropsychiatric efficacy outcomes compared to placebo. NPI total score was lower in 4 studies (SMD −0.31; 95% CI −0.47 to −0.15; p not individually reported for this outcome but CI excludes null). CMAI-SF agitation scores were lower in 3 studies (SMD −0.40; 95% CI −0.69 to −0.10; p=0.009). NPI agitation/aggression domain scores were lower in 3 studies (SMD −0.47; 95% CI −0.69 to −0.25; p<0.001). All effect sizes are in the small-to-moderate range per conventional SMD thresholds (Cohen's d 0.2–0.5); NNT is not calculable from SMD alone without raw score distributions. MMSE findings were not supportive of consistent cognitive benefit or harm. The principal adverse event was somnolence (including sedation and lethargy): RR 2.25 (95% CI 1.43–3.54; p=0.0005), representing the only adverse event consistently observed at higher rates in the cannabinoid group. Absolute risk difference for somnolence was not reported. Bayesian and Trial Sequential Analyses were performed but specific Bayes factors, posterior probabilities, and TSA monitoring boundary data are not reported in available sources.
Practice implications
This meta-analysis provides the most methodologically rigorous pooled estimate to date showing that cannabinoid-based therapies are associated with small-to-moderate reductions in agitation and neuropsychiatric symptom burden in moderate-to-severe AD versus placebo. The findings are statistically significant across three distinct validated instruments, which adds some convergent validity. However, several critical limitations preclude practice change. First, the total sample of 221 patients across 7 studies is very small for a meta-analysis; such small pools are disproportionately sensitive to overestimation of treatment effects and imprecise heterogeneity estimation. Second, trial durations of 2–26 weeks are insufficient to characterize durability of benefit or long-term safety in a chronically ill, elderly population. Third, the five formulations studied (dronabinol, nabiximols, nabilone, THC oil, THC-CBD extract) are not pharmacologically interchangeable, and pooling them assumes class-level equivalence that has not been established. Fourth, a critical confound exists: lower agitation scores may reflect genuine symptom relief, cannabinoid-mediated sedation, or both — the study design cannot distinguish these mechanisms, and somnolence was doubled in the treatment group (RR 2.25). Fifth, no active comparator arm was included, so relative efficacy against currently used agents (e.g., quetiapine, valproate) is unknown. The MMSE findings did not demonstrate consistent cognitive worsening, which is reassuring, but absence of signal in a small study should not be equated with cognitive safety. Findings apply most directly to adults with moderate-to-severe AD; patients with mild AD, those on complex polypharmacy, or those with baseline fall risk or hepatic impairment were not adequately represented. Note: only an abstract and secondary media reports were available for review; detailed methodology, heterogeneity statistics (I²), risk-of-bias data, and complete TSA/Bayesian outputs could not be independently verified. __LIMITATIONS__ Very small total sample (n=221 across 7 studies, only 6 of which were RCTs); highly susceptible to overestimation of treatment effects and imprecise heterogeneity estimates,Heterogeneous cannabinoid formulations, doses, and routes of administration pooled under a single class assumption without formulation-specific subgroup power,Short and highly variable trial durations (2–26 weeks); insufficient to assess durability of benefit or long-term safety in a chronic geriatric condition,Absence of active comparator arm; relative efficacy versus existing pharmacological agents for AD agitation (e.g., antipsychotics, mood stabilizers) cannot be determined,Somnolence signal (RR 2.25) confounds interpretation of agitation reduction: it is unclear whether lower agitation scores reflect true symptom improvement or sedation,Heterogeneous outcome reporting and variable baseline severity across studies limit poolability,No caregiver burden, institutionalization, quality of life, or functional independence outcomes reported,Generalizability limited: patients with mild AD, significant comorbidities, high fall risk, or complex polypharmacy were likely underrepresented,Funding source not reported; potential for publication bias in a small literature base,Full-text analysis not possible; summary based on abstract, secondary reporting, and partial data — Bayesian and TSA outputs, I² values, and risk-of-bias assessments not independently verifiable __BOTTOM_LINE_CATEGORY__ promising_but_preliminary __BOTTOM_LINE__ Cannabinoid-based therapies demonstrated small-to-moderate, statistically significant reductions in agitation and neuropsychiatric symptom scores versus placebo in moderate-to-severe AD across three validated instruments, with somnolence as the predominant and consistently doubled safety signal. These findings are hypothesis-generating but not practice-changing: the evidence base rests on only 221 patients across heterogeneous formulations and short trials insufficient to disentangle symptom relief from sedation. Larger, formulation-specific, active-controlled RCTs with longer follow-up and standardized safety monitoring are required before cannabinoids can be recommended as a routine therapeutic option for AD-related agitation.