Cannabinoids for the treatment of dementia

Abstract

Background

Dementia is a common chronic condition, mainly affecting older adults, characterised by a progressive decline in cognitive and functional abilities. Medical treatments for dementia are limited. Cannabinoids are being investigated for the treatment of dementia.

Objectives

To determine the efficacy and safety of cannabinoids for the treatment of dementia.

Search methods

We searched ALOIS ‐ the Cochrane Dementia and Cognitive Improvement Group’s Specialised Register ‐ on 8 July 2021, using the terms cannabis or cannabinoid or endocannabinoid or cannabidiol or THC or CBD or dronabinol or delta‐9‐tetrahydrocannabinol or marijuana or marihuana or hashish. The register contains records from all major healthcare databases (the Cochrane Library, MEDLINE, Embase, PsycINFO, CINAHL, LILACS), as well as from many clinical trials registries and grey literature sources.

Selection criteria

We included all randomised controlled trials (RCTs) of cannabinoids for the treatment of dementia. We included participants of any age and of either sex with diagnosed dementia of any subtype, or with unspecified dementia of any severity, from any setting. We considered studies of cannabinoids administered by any route, at any dose, for any duration, compared with placebo, no treatment, or any active control intervention.

Data collection and analysis

Two review authors independently screened and selected studies for inclusion, extracted data, and assessed the risk of bias in included studies. When necessary, other review authors were involved in reaching consensus decisions. We conducted meta‐analyses using a generic inverse variance fixed‐effect model to derive estimates of effect size. We used GRADE methods to assess our confidence in the effect estimates.

Main results

We included four studies (126 participants) in this review. Most participants had Alzheimer's disease; a few had vascular dementia or mixed dementia. Three studies had low risk of bias across all domains; one study had unclear risk of bias for the majority of domains. 

The included studies tested natural delta‐9‐tetrahydrocannabinol (THC) (Namisol) and two types of synthetic THC analogue (dronabinol and nabilone). Three trials had a cross‐over design. Interventions were applied over 3 to 14 weeks; one study reported adverse events over 70 weeks of follow‐up. One trial was undertaken in the USA, one in Canada, and two in The Netherlands. Two studies reported non‐commercial funding, and two studies were conducted with the support of both commercial and non‐commercial funding.

Primary outcomes in this review were changes in global and specific cognitive function, overall behavioural and psychological symptoms of dementia (BPSD), and adverse events.

We found very low‐certainty evidence suggesting there may be little or no clinically important effect of a synthetic THC analogue on cognition assessed with the standardised Mini‐Mental State Examination (sMMSE) (mean difference (MD) 1.1 points, 95% confidence interval (CI) 0.1 to 2.1; 1 cross‐over trial, 28 participants). 

We found low‐certainty evidence suggesting there may be little or no clinically important effect of cannabinoids on overall behavioural and psychological symptoms of dementia assessed with the Neuropsychiatric Inventory (or its modified nursing home version) (MD ‐1.97, 95% CI ‐3.87 to ‐0.07; 1 parallel group and 2 cross‐over studies, 110 participants).

All included studies reported data on adverse events. However, the total number of adverse events, the total numbers of mild and moderate adverse events, and the total number of serious adverse events (SAEs) were not reported in a way that permitted meta‐analysis.  There were no clear differences between groups in numbers of adverse events, with the exception of sedation (including lethargy), which was more frequent among participants taking nabilone (N = 17) than placebo (N = 6) (odds ratio (OR) 2.83, 95% CI 1.07 to 7.48; 1 cross‐over study, 38 participants). We judged the certainty of evidence for adverse event outcomes to be low or very low due to serious concerns regarding imprecision and indirectness.

Authors' conclusions

Based on data from four small, short, and heterogeneous placebo‐controlled trials, we cannot be certain whether cannabinoids have any beneficial or harmful effects on dementia. If there are benefits of cannabinoids for people with dementia, the effects may be too small to be clinically meaningful. Adequately powered, methodologically robust trials with longer follow‐up are needed to properly assess the effects of cannabinoids in dementia.

Plain language summary

Cannabinoids for the treatment of dementia

Background

Dementia is the name for a range of conditions associated with an ongoing loss of thinking ability, memory, and other mental abilities. Medical treatments for dementia are limited and confer modest benefits. Many medications and non‐pharmacological interventions are used for behavioural and psychological symptoms of dementia. However,  there are continuing problems with their lack of efficacy, safety, and feasibility. Accordingly, new, safe, and more effective treatments are needed for dementia and its associated symptoms. 

Our review question

The cannabinoids are one potential agent under investigation for the treatment of dementia. The purpose of this systematic review was to investigate whether cannabinoids could help people with dementia, and whether they have any potential harmful effects.

What we did

We searched databases of scientific studies to find studies that had to randomly decide whether people would be treated with cannabinoids or a comparator. We combined the results of included studies to estimate the effects of cannabinoids. We also assessed how well these studies were conducted and how credible their results were.

What we found

We searched for relevant studies that had been published up to June 2020. We found four trials that met the inclusion criteria for this review. A total of 126 people were included in those four trials. Most participants had Alzheimer's disease, and a few had vascular dementia or mixed dementia.

One trial was undertaken in the USA, one in Canada, and two in The Netherlands. Trials used different types of cannabinoids. Studies reported that cannabinoids have little or no effect on memory and thinking. They reported different results regarding overall behavioural and psychological symptoms of dementia, based on the types of individuals who were reporting data. Family caregivers did not report a beneficial effect of cannabinoids on behaviour and psychological symptoms, but nursing staff reported improvement of symptoms among participants receiving cannabinoids. Harms were reported in all studies, among both participants taking cannabinoids and those taking placebo. However, we could not combine the total number of harmful events due to problems with data reporting. There was no significant difference between participants taking cannabinoids and those given placebo in the total numbers of harmful events belonging to nervous system disorders, psychiatric disorders, and gastrointestinal disorders. Sedation (including lethargy) was more frequent in participants taking cannabinoids, but these results were uncertain. We have very low confidence in these results because studies included a small number of participants, there were differences between the studies, and their results were uncertain.

Our conclusions

Based on data from four small trials of short duration, it is uncertain whether cannabinoids have any beneficial or harmful effects on dementia, compared to placebo. Even if the benefit reported in these studies is real, the effect was modest and may not be important to people living with dementia. Furthermore, available studies were very short, with efficacy examined over 3 to 14 weeks, and one study did not report its methods and results completely. A large, well‐conducted study is needed to understand better if cannabinoids are a useful treatment for people living with dementia.

Summary of findings

Summary of findings 1. Comparison: cannabinoids compared to placebo for the treatment of dementia.

Comparison: cannabinoids compared to placebo for the treatment of dementia
Patient or population: diagnosis of dementia (dementia in Alzheimer’s disease, vascular dementia, or mixed dementia; any severity Setting: hospital, nursing home, outpatient treatment Intervention: all were tetrahydrocannabinol (THC), administered orally Comparison: placebo
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with placebo	Risk with comparison: cannabinoids
Changes in global and specific cognitive functions, sMMSE scale (follow‐up: 6 weeks each treatment period); sMMSE score: 0 to 30, with lower number of points indicating more severe dementia		MD 1.1 higher (0.1 higher to 2.1 higher)	‐	39 (1 RCT)	⊕⊝⊝⊝ VERY LOW 1 2	This was a cross‐over trial with 39 participants. For sMMSE, it has been proposed that the minimal clinically important difference (MCID) is 1.4 in dementia patients (Howard 2011)
Overall behavioural and psychological symptoms of dementia, NPI  and NPI‐NH total score (follow‐up: 3 to 14 weeks); NPI total score: 0 to 144, with 0 meaning no symptoms, and higher number of points indicating more severe symptoms		MD 1.97 lower (3.87 lower to 0.07 lower)	‐	111 (3 RCTs)	⊕⊕⊝⊝ LOW 3 4	This analysis included 2 cross‐over studies and 1 parallel‐group trial. For NPI‐MH, the MCID of 8 points in dementia patients has been suggested, but based on an estimate from 1 study and in a particular stage of dementia (Howard 2011)
Agitated or aggressive behaviours in dementia patients, measured on the Cohen‐Mansfield Agitation Inventory (CMAI) Scale (range 29 to 203, with 29 meaning no symptoms, and higher number of points indicating more severe symptoms)		MD 2.35 lower (4.10 lower to 0.60 lower)	‐	100 (3 RCTs)	⊕⊕⊝⊝  LOW 3 4	This analysis included 2 cross‐over studies and 1 parallel‐group trial
Adverse events ‐ nervous system disorders ‐ total nervous system disorders (follow‐up: 5 weeks)	Study population		OR 0.71 (0.23 to 2.18)	50 (1 RCT)	⊕⊕⊝⊝ LOW 2 5
	500 per 1000	415 per 1000 (187 to 686)
Adverse events ‐ psychiatric disorders ‐ total psychiatric disorders (follow‐up: 5 weeks)	Study population		OR 2.26 (0.57 to 9.02)	50 (1 RCT)	⊕⊕⊝⊝ LOW 2 5
	154 per 1000	291 per 1000 (94 to 621)
Adverse events ‐ gastrointestinal disorders ‐ total gastrointestinal disorders (follow‐up: 5 weeks)	Study population		OR 2.40 (0.40 to 14.49)	50 (1 RCT)	⊕⊕⊝⊝ LOW 2 5
	77 per 1000	167 per 1000 (32 to 547)
Adverse events ‐ general ‐ total general adverse events (follow‐up: 5 weeks)	Study population		OR 0.70 (0.11 to 4.58)	50 (1 RCT)	⊕⊕⊝⊝ LOW 2 5
	115 per 1000	84 per 1000 (14 to 374)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; OR: odds ratio; RR: risk ratio.
GRADE Working Group grades of evidence. High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

¹ Downgraded two levels for indirectness: very short study duration (14‐week randomized double‐blind cross‐over trial compared nabilone to placebo, 6 weeks each, with a 1‐week washout between phases; effect on cognition unlikely to be detected.

² Downgraded one level for imprecision: only one small study.

³ Downgraded one level for inconsistency: large statistical heterogeneity.

⁴ Downgraded one level for imprecision: multiple small trials.

⁵ Downgraded one level for imprecision: confidence interval crosses the line of no effect.

Background

Description of the condition

Dementia is a common chronic condition mainly affecting older adults and characterised by a progressive decline in cognitive and functional abilities. The most common forms of dementia include Alzheimer’s disease (AD) (60% to 70% of cases), vascular dementia (VaD), dementia with Lewy bodies (DLB), dementia in Parkinson's disease (PDD), and frontotemporal dementia (FTD). The boundaries between different subtypes of dementia are indistinct, and mixed forms often co‐exist (WHO 2013).

It has been estimated that 47 million people worldwide were living with dementia in 2016, and the number was projected to increase to more than 131 million in 2050 due to population ageing (ADI 2016). This disabling condition brings with it a significant burden to individuals and their carers, as well as a large financial burden for the health system (Standfield 2018), thus driving the need to identify effective therapeutic interventions.

Medical treatments for dementia are limited, with an emphasis on managing symptoms. Licensed medications are available only for dementia due to AD and PDD, and these have only modest benefit for cognitive symptoms. At least half of patients with dementia will also experience behavioural and psychological symptoms (BPSD) such as agitation, aggression, psychosis, and circadian rhythm disturbances. These symptoms lead to significant caregiver stress (Rabins 1982), are distressing for the patient, and often precipitate placement in residential or nursing homes (Steele 1990). Antipsychotic drugs are widely used to treat BPSD but have only modest efficacy (Ballard 2006; Schneider 2006). Use of these drugs in dementia is associated with serious side effects including increased risk of cerebrovascular adverse events and death (FDA 2005; MHRA 2004; Schneider 2005). A range of pharmacological and non‐pharmacological interventions are used for BPSD, but there are continuing problems with their lack of efficacy, safety, and feasibility. Accordingly, new, safe, and more effective treatments for dementia and its associated symptoms.

Description of the intervention

The cannabinoids are one potential agent under investigation for the treatment of dementia. Multiple studies have indicated that patients have positive attitudes towards medical cannabis (Banwell 2016; Gazibara 2017). There are three general classes of cannabinoids, including herbal cannabinoids (phytocannabinoids) that are derived from the cannabis plant (Cannabis sativa), endogenous cannabinoids that are produced in bodies of humans and animals, and synthetic cannabinoids that are produced in a laboratory. Cannabis, also known as marijuana, is a preparation of the Cannabis plant. It is one of the most popular recreational drugs; an estimated 183 million users used cannabis globally in 2015, which roughly corresponds to 3.8% of the global population (UNODC 2017). The use of cannabis is illegal in most countries. Although the general public may perceive cannabis as the least harmful illicit drug, there has been a noticeable increase over the past decade in the number of patients seeking treatment for disorders related to cannabis use (UNODC 2017).

Some countries have recently legalised medicinal‐grade cannabis for chronically ill patients. Nausea and vomiting due to chemotherapy, appetite stimulation in HIV/AIDS, chronic pain, spasticity due to multiple sclerosis or paraplegia, depression, anxiety disorder, sleep disorder, psychosis, glaucoma, and Tourette syndrome are some indications for its legal use. Anxiety, drowsiness, euphoria, dry mouth, psychosis, dizziness, and diarrhoea are known adverse effects (Whiting 2015).

The first cannabinoids to be identified were the main psychoactive compound delta‐9‐tetrahydrocannabinol (THC) and the non‐psychoactive compound cannabidiol (CBD), although there are thought to be numerous other cannabinoids, some of which may modulate the response to THC (Iversen 2000). Research for medical use of cannabinoids has resulted in development and marketing of the synthetic cannabinoids dronabinol and nabilone. Dronabinol may be taken orally or via local/topical, transdermal, sublingual, or inhaled modes of administration. Nabilone is taken orally as a capsule. Daily dosage and route of administration depend on the indication (Howard 2013). Nabilone and dronabinol are used as interventions for reducing vomiting associated with chemotherapy. A combination of cannabidiol and THC (Sativex) has been approved in several countries for treating spasticity in multiple sclerosis, and it is being studied for cancer pain. One cannabidiol drug (Epidiolex) has been studied for use in epilepsy. Many ongoing studies are exploring therapeutic targets for both cannabinoid receptor agonists and antagonists (Kaur 2016).

Cannabinoids exert their effect by acting at two specific cannabinoid receptors, CB1 and CB2, in the endogenous cannabinoid system (Howlett 2002; Matsuda 1990). CB1 receptors are found throughout the central nervous system, particularly in the hippocampus, basal ganglia, and cerebellum. In contrast, CB2 receptors are expressed in peripheral tissues, especially on white blood cells, and are much less widespread in the central nervous system (see Campbell 2007 for a review). Several studies have identified CB2 receptors on brainstem neurons ‐ Van Sickle 2005 ‐ and cerebellar neurons ‐ Onaivi 2006 ‐ but their role is not yet understood.

How the intervention might work

Several demonstrated neurobiological effects of cannabinoids could be relevant for the treatment of dementia. The main function of the endogenous cannabinoid system (ECS) is thought to be regulation of synaptic transmission (Baker 2003); this process can be disordered in many neurological conditions including dementia. The ECS consists of endogenous cannabinoids (endocannabinoids), cannabinoid receptors, and enzymes responsible for synthesis and degradation of endocannabinoids (Lu 2016). The best characterised endogenous cannabinoids are anandamide (arachidonoyl ethanolamide) and 2‐arachidonoyl glycerol (2‐AG); some less well characterised additional endogenous substances such as virodhamine and 2‐arachidonoyl glycerol ether also belong to the repertoire of endocannabinoids. Endocannabinoids exert their actions mainly by acting on cannabinoid receptors CB1 and CB2 (Gowran 2011); other receptors such as peroxisome proliferator activated receptors (PPARs) and transient receptor potential (TRP) channels also mediate certain endocannabinoid actions, especially of the acylethanolamides. Relevant enzymes are fatty acid amino hydrolase (FAAH) involved in degradation of anandamide, monoacylglycerol lipase (MGL), and alpha/beta domain hydrolases 6 and 12 (ABHD6 and 12) involved in degradation of 2‐AG (Lu 2016).

CB1 receptors are found mainly in the central and peripheral nervous system, where they usually mediate inhibition of ongoing release of different neurotransmitters (Szabo 2005). Activation of CB1 affects cognition and memory, alters control of motor function, and induces signs of analgesia (Pertwee 2010). CB1 receptors regulate processes such as excessive glutamate production and subsequent oxidative stress, which can damage neurons and lead to neurodegeneration (Grundy 2002). They are also found in peripheral tissues where they have a role in energy balance and metabolism (Silvestri 2013). 

CB2 receptors modulate immune cell migration and cytokine release, as most are located in the immune cells (Pertwee 2005); in the central nervous system, they are mainly located in the microglia ‐ Cabral 2009 ‐ and in some neurons ‐ Brusco 2008, Onaivi 2008 ‐ where they are connected with facilitation of neuronal survivor (Viscomi 2009).

Some evidence suggests that CB2 receptors may be involved in neuroprotection by reducing neuroinflammation (Ehrhart 2005). Neurodegeneration is a feature common to the various types of dementia; therefore the neuroprotective effects of cannabinoids may be beneficial in slowing progression of these diseases.

Some trial authors have reported a significant reduction in CB1 levels in cortical areas and neurons distant from senile plaques (Solas 2013); others have indicated that there are changes in the expression, distribution, and availability of CB1 in AD (Ahmad 2014; Mulder 2011). According to Solas and colleagues, CB2 levels are significantly increased in the brains of people with AD, mainly on the microglia around the senile plaques (Solas 2013). Besides CB1 and CB2, cannabinoids can bind to other types of receptors, such as GPR55, peroxisome proliferator‐activated receptors PPARα and PPARγ, and transient receptor potential vannilloid‐1 (TRPV1) channels (Maccarrone 2010; Pertwee 2010).

Cannabinoids may have more specific effects in Alzheimer's disease pathology, as they can reduce excitotoxicity, mitochondrial dysfunction, oxidative stress, neuroinflammation, and the formation of amyloid plaques and neurofibrillary tangles (Ahmed 2015; Aso 2014). Several studies have shown the protective effect of cannabinoids against amyloid‐β peptide and tau phosphorylation (reviewed in: Aso 2014), which are the neuropathological hallmarks of AD.

THC diminishes acetylcholinesterase‐induced amyloid beta‐peptide aggregation ‐ the key pathological marker of AD (Ahmed 2015; Eubanks 2006). It has also been reported that THC competitively inhibits the enzyme acetylcholinesterase (AChE) ‐ an action similar to that of anti‐dementia drugs like donepezil (Ahmed 2015). Another study investigated effects of cannabinoids in rats injected with amyloid beta‐peptide to model AD. Intracerebroventricular administration of a synthetic cannabinoid (WIN55,212‐2) to these rats led to prevention of their cognitive deficit and decreased neurotoxicity (Janefjord 2014; Ramirez 2005). These studies suggest that cannabinoids could interrupt the disease process as well as treat symptoms in AD. Endocannabinoids (anandamide, 2‐AG, noladin ether) seem to increase the viability of neurons after exposure to toxic Aβ species (Chen 2011; Harvey 2012). Similar effect was reported with exogenous cannabinoids such as CBD (Janefjord 2014), arachidonyl‐2‐chloroethylamide (Aso 2012), JWH‐015, JWH‐133, and HU‐210 (Ramirez 2005). Several studies demonstrated positive results in prevention of memory deficits in Aβ‐injected rats and mice for exogenous cannabinoids (Aso 2013; Martin‐Moreno 2012; Wu 2013).

These studies indicate that the disease‐modifying action of cannabinoids is most likely. However, cannabinoids also have known symptomatic effects, which may be of benefit in dementia. The most common neuropsychiatric symptoms (NPSs) in dementia, including depression, anxiety, agitation, aggression, wandering, pacing, sleep disorders, psychosis, and eating disorders, are associated with more rapid dementia progression and higher healthcare costs (Beeri 2002; Tschanz 2013). To this day, no drugs have been approved by the US Food and Drug Administration (FDA) for treatment of NPSs associated with AD dementia (Ahmed 2015), and in the European Union and Australia, only the antipsychotic risperidone is indicated for short‐term management of severe aggression in patients with AD, with unsuccessful non‐pharmacological methods (Panza 2015). A recent review indicated that findings from six studies show significant benefit from synthetic cannabinoids dronabinol or nabilone for agitation and aggression; however conclusions were limited by small sample size, short trial duration, and lack of placebo control in some studies (Liu 2015). In another randomised controlled trial (van den Elsen 2015), oral THC 4.5 mg daily showed no benefit for treatment of NPSs, but it was well tolerated, which allows further study on whether higher doses would be more efficient. An open‐label pilot study evaluated the effect of dronabinol for treatment of NPSs in dementia and showed that it significantly improved nocturnal motor activity and behaviour (Walther 2006). A recent retrospective systematic chart review suggested that administration of dronabinol improved sleep duration, food consumption, and agitation (Woodward 2014).

Why it is important to do this review

With increased life expectancy and increased numbers of people aged 60 years and older, the number of new cases of dementia is projected to rise. The need to offer effective and safe interventions to people with dementia is growing. Because firm evidence of efficacy and safety of cannabinoids in this vulnerable patient group is lacking, a systematic review can help inform decisions of healthcare workers, researchers, politicians, and other public health decision‐makers.

Objectives

To determine the efficacy and safety of cannabinoids for the treatment of dementia.

Methods

Criteria for considering studies for this review

Types of studies

We included all randomised controlled trials (RCTs) of cannabinoids for the treatment of dementia.

Types of participants

We included participants of any age and of either sex with diagnosed dementia of any subtype or with unspecified dementia of any severity from any setting. The diagnosis should be made using internationally recognised criteria including Diagnostic and Statistical Manual of Mental Disorders V (DSM V) or previous editions of DSM (APA 2013), the International Classification of Diseases 10 (ICD 10) or previous editions of ICD (WHO 2010), National Institute of Neurological Disorders and Stroke (NINDS)‐Association Internationale pour la Recherche et l'Enseignement en Neurosciences (AIREN) Criteria for the Diagnosis of Vascular Dementia (Roman 1993), or National Institute of Neurological and Communicative Diseases and Stroke (NINCDS)‐Alzheimer's Disease and Related Disorders Association (ADRDA) Alzheimer's criteria (McKhann 2011). Studies that included only a subset of relevant participants were included only if data for the population of interest were reported separately.

Types of interventions

Experimental interventions: cannabinoids administered by any route, at any dose, for any duration.

Control interventions: placebo, no treatment, or any active control intervention.

Types of outcome measures

Primary outcomes

1. Changes in global and specific cognitive functions, measured by any validated cognitive scale covering multiple cognitive domains or single cognitive domains (e.g. memory, executive function), such as the Standardised Mini‐Mental State Examination (sMMSE) (Folstein 1975)

2. Overall behavioural and psychological symptoms of dementia (BPSD), measured with any validated instrument (e.g. the Neuropsychiatric Inventory (NPI) (Cummings 1994), the Alzheimer's Disease Assessment Scale (ADAS)) (Rosen 1984)

3. Adverse events

Secondary outcomes

1. Changes in functional outcomes, such as activities of daily living (ADLs), measured by validated tools such as:

   1. Alzheimer’s Disease Activities of Daily Living International Scale (ADCS‐ADL) (Galasko 1997); and

   2. Gottries‐Brane‐Steen‐Skala, ADL subscale (GBS‐ADL) (Brane 2001)

2. Overall dementia severity measured by validated tools such as:

   1. Clinical Dementia Rating Scale‐Sum of Boxes (CDR‐SOB) (O'Bryant 2008); and

   2. Alzheimer's Disease Cooperative Study‐Clinical Global Impression of Change (CIBIC‐Plus) (Schneider 1997)

3. Objective sleep outcomes measured with polysomnography or actigraphy

   1. Total nocturnal sleep time (TNST; i.e. total time spent asleep between 8.00 PM and 8.00 AM)

   2. Sleep efficiency (%; i.e. TNST/time in bed x 100)

   3. Nocturnal time awake (WASO; after sleep onset and before final awakening)

   4. Number of nocturnal awakenings

   5. Sleep latency

   6. Ratio of daytime sleep to night‐time sleep, or of night‐time sleep to total sleep over 24 hours

4. Changes in appetite

   1. Change in body weight (kg)

   2. Carer ratings of patient's anorexia or change in appetite, or both

5. Agitated or aggressive behaviours

   1. Quantitative observational tools for measuring frequency of occurrence of wandering, agitation, and general restlessness

   2. Standardised tests such as Cohen‐Mansfield Agitation Inventory (Cohen‐Mansfield 1989)

6. Mood, measured with any validated tool

7. Carer ratings of patient's sleep using sleep diaries or validated observer scales

8. Quality of life

9. Any other symptoms associated with dementia (e.g. alterations in circadian rhythm)

10. Caregiver burden and caregiver quality of life

11. Treatment or research discontinuation/dropout (as measures of acceptability)

12. Mortality

We did not include biomarker outcomes.

Search methods for identification of studies

Electronic searches

We searched ALOIS (www.medicine.ox.ac.uk/alois) ‐ the Cochrane Dementia and Cognitive Improvement Group’s Specialised Register, on 8 July 2021. The search terms (separated by OR) used were cannabis, cannabinoid, endocannabinoid, cannabidiol, THC, CBD, dronabinol, delta‐9‐tetrahydrocannabinol, marijuana, marihuana, and hashish.

ALOIS is maintained by the Information Specialists of the Cochrane Dementia and Cognitive Improvement Group and contains studies in the areas of dementia prevention, dementia treatment, and cognitive enhancement in healthy individuals. Studies are identified from:

1. monthly searches of a number of major healthcare databases: MEDLINE, Embase, CINAHL, PsycINFO, and LILACS;

2. monthly searches of a number of trial registers: ISRCTN (Current Controlled Trials); UMIN (Japan's Trial Register); the World Health Organization (WHO) portal (ICTRP) (which covers ClinicalTrials.gov; ISRCTN; the Chinese Clinical Trials Register; the German Clinical Trials Register; the Iranian Registry of Clinical Trials, and the Netherlands National Trials Register, plus others);

3. quarterly search of the Cochrane Library’s Central Register of Controlled Trials (CENTRAL); and

4. six‐monthly searches of a number of grey literature sources: ISI Web of Knowledge Conference Proceedings

Details of the search strategies used for retrieval of reports of trials from healthcare databases, CENTRAL, and conference proceedings can be viewed in the ‘Methods used in reviews’ section within editorial information about the  Dementia and Cognitive Improvement Group.

Additional searches were performed for many of the sources listed above to cover the time frame from the last searches performed for ALOIS to ensure that the search for the review was as up‐to‐date and as comprehensive as possible. The search strategies used can be seen in Appendix 1.

Searching other resources

Previously, we searched a number of other resources including the Google search engine and the Norml website http://www.norml.org/index.cfm to identify other relevant references. In addition, we contacted the first authors of two relevant trials to request details of any unpublished or current studies that might meet inclusion criteria for this review. We also examined the reference lists of retrieved articles to look for additional trials for inclusion.

Data collection and analysis

Selection of studies

We managed all references retrieved by the searches using EndNote X5 software (EndNote 2011). The Systematic Review Assistant‐Deduplication Module (SRA‐DM) was used to identify and remove duplications of the same references (Rathbone 2015).

Two review authors independently screened titles and abstracts from all bibliographic records retrieved via the literature search to identify eligible studies. If it appeared from title and abstract that a study might be eligible, we obtained a full text of the report to make a decision. If available, we also obtained all errata and supplementary data for all full texts of studies eligible for inclusion. If necessary, we planned to translate full texts of reports that were not published in English, by employing a translation service. However, we did not find such studies.

When we found multiple reports of the same study, we linked them together. Two review authors independently evaluated the full texts of relevant articles according to eligibility criteria. The review authors were not blinded to study data. We resolved possible disagreements via discussion and, if necessary, through involvement of a third review author. We listed justifications for exclusion of articles that were retrieved in full text. We documented the study selection process as suggested in the PRISMA statement (Liberati 2009).

Data extraction and management

Two review authors independently extracted relevant data from the included studies. We resolved potential discrepancies via discussion and involvement of a third review author if necessary. In case of language ambiguity, we contacted researchers in the field familiar with the language in question. We used an electronic data extraction form, including source, eligibility, methods, participants, interventions, comparators, outcomes, results, and miscellaneous notes, according to the Cochrane Handbook for Systematic Reviews of Interventions (Chapter 7.3; Higgins 2011). We also collected information on details of funding sources, declarations of interest of primary investigators, and methods used to control possible conflicts of interest. Two review authors independently piloted the form using two studies. The data extraction form was adapted thereafter, if necessary.

For continuous data in a parallel‐group trial, we extracted the mean value of the outcome measurement in each group at each time point, the standard deviations (SDs), and the number of participants used to measure the outcome for each group. As the authors of this trial also reported estimates of overall mean differences over Days 14 and 21, which were based on linear mixed‐model analysis for repeated measures with correction for NPI score at baseline ‐ center, Clinical Dementia Rating stage, sex, current opioid use, week, and use of a random intercept; we also extracted these data as generic inverse variance data and used them in analysis.

For dichotomous outcomes from the same study, we extracted the number of events and the total number of participants for each outcome group at each time point.  

For cross‐over trials, we entered data as generic inverse variance data using estimates from paired analyses. For the two time‐to‐event outcomes ‐ all‐cause research discontinuation/dropout and all‐cause mortality ‐ for which no analysis was reported in the studies, we extracted the number of events and the total number of participants for each treatment period. 

If we did not find the necessary data, we tried to complete them with the help of the primary study authors (see the section Dealing with missing data). 

One review author entered the data into Review Manager 5 (RevMan 2014), and another review author checked the data for accuracy.

When we found published protocols of eligible studies, we extracted data from ongoing studies, including study name, methods, participants, interventions, outcomes, starting date, contact information, and notes.

In the review protocol, for data published only in figures/graphs, we planned to request data from the corresponding authors or, alternatively, to extract them using Plot Digitizer software (Jelicic Kadic 2016; Vucic 2015). However, during the conduct of the review, we did not encounter data that had to be extracted from figures.

Assessment of risk of bias in included studies

Two review authors independently assessed the risk of bias for each study, using the Cochrane 'Risk of bias' tool, according to the Cochrane Handbook for Systematic Reviews of Interventions (Chapter 8.5; Higgins 2011). For all included studies, we described the risk of bias both in tables and narratively in text. We also provided an overall judgement about included studies in the 'Risk of bias' tables and the 'Risk of bias' charts. We analysed whether there was any information addressing the appropriateness of methods used to prevent undue industry influence during the clinical trial process.

Measures of treatment effect

We used mean differences (MDs) or standardised mean differences (SMDs) with 95% confidence intervals (CIs) for continuous outcomes, and odds ratios (ORs) with 95% CIs for analysis of dichotomous outcomes.  We considered ordinal outcomes only if we could justifiably treat them as a continuous variable, or if they could be sensibly dichotomised by combining adjacent categories. Given that there are no definitive guidelines for handling these measurements, we reported on our decision, which was reached in a discussion that involved at least two review authors. 

Unit of analysis issues

The unit of analysis in parallel‐group and cross‐over trials was an individual, with an individual in a cross‐over trial observed multiple times. We accounted for any unit of analysis errors originating from the study design by including only results on paired data analyses from cross‐over trials. We did not identify any trial with multiple treatment arms.

Dealing with missing data

If data from a study were missing and could not be calculated using available data and statistics, we tried to contact the trial authors to complete the data. To contact study authors, we made at least two contact attempts via email or other potential means of contact over six weeks, checking for alternate contact information if the first attempt failed. In the case we were unable to retrieve the complete data, we reported this in the 'Risk of bias' assessment and addressed missing outcomes and summary data as a source of bias in data analysis.

We carried out analyses on an intention‐to‐treat (ITT) basis for all outcomes, as far as possible. 

Assessment of heterogeneity

Whenever we had two or more studies in analysis, we measured and described heterogeneity of the treatment effect between trials using the I² statistic and the P value from the corresponding Chi² test. We also double‐checked this visually by using forest plots. For assessment of clinical heterogeneity, we analysed the data extraction tables and considered the data for between‐study variability with respect to participants, interventions, and outcome measurements.

Assessment of reporting biases

To minimise reporting bias, we planned in the protocol to include both published and unpublished trials. For detecting possible reporting bias, we planned to use a funnel plot and Egger's test for asymmetry, if enough (more than 10) studies were available (Egger 1997). However, due to insufficient numbers of trials identified, we did not make this assessment for possible reporting bias.

When applicable, we compared conference abstracts and available trial protocols of included studies with published data.

Data synthesis

We used Review Manager 5 software to perform all statistical analyses (RevMan 2014).

We used meta‐analysis for combining data if (i) at least two studies reported an estimated treatment effect, (ii) included studies appeared to have similar characteristics, (iii) studies had the same outcome measures, and (iv) each study reported the necessary data.

For data synthesis, we considered all listed primary and secondary outcomes. Except for the two outcomes discussed below, we analysed all efficacy study data using the generic inverse variance fixed‐effect model to determine overall weighted treatment effects and their 95% CIs. This included data for continuous outcomes that were informed by a single parallel‐group trial, as for these outcomes, study authors reported an estimate of overall mean differences over several time points. For the two rare, time‐to‐event outcomes ‐ all‐cause research discontinuation/dropout and all‐cause mortality ‐ no analysis was reported in studies observing the events. We, therefore, analysed them as binary and used a conservative approach to interpret events from different treatment periods as parallel‐group events using the Mantel‐Haenszel method and OR as an effect measure. If types of participants, interventions, comparisons, and outcome measures used were very different between studies, we used only narrative interpretative synthesis of data for individual studies separately.

Even though all studies reported data on adverse events, due to problems with data presentation, we could not conduct a meta‐analysis for a total number of adverse events, or a total number of serious adverse events. We used data from the only parallel‐group trial to present data for specific types of adverse events ‐ nervous system disorders, psychiatric disorders, gastrointestinal disorders, and general symptoms ‐ that were reported previously as the most frequent adverse events of cannabinoids (Wang 2008). For this one parallel‐group trial, we presented ORs as measures of treatment effect using the Mantel‐Haenszel method.

Subgroup analysis and investigation of heterogeneity

We aimed to conduct the following subgroup analyses, if data are available.

1. Dose of intervention.

2. Type of dementia.

3. Stage of dementia, differentiating very mild, mild, moderate, and severe dementia, as defined by validated tools such as the CDR‐SOB (O'Bryant 2008).

We were planning to perform tests for heterogeneity using the Chi² test and the I² statistics within each of these groups. For each subgroup, if sufficient data are available, we planned to perform a meta‐analysis using either a random‐effects model or a fixed‐effect model, depending on the presence of statistical heterogeneity. However, due to insufficient data, we were unable to perform these subgroup analyses.

We performed a post‐hoc subgroup analysis on overall behavioural and psychological symptoms of dementia. We identified one study as a separate subgroup because of differences among participants (more severely ill patients), interventions (synthetic THC analogue instead of oral THC), and outcome measures (NPI‐NH versus NPI).

Sensitivity analysis

In the review protocol, we planned to conduct sensitivity analyses to explore the robustness of analyses. For meta‐analyses, we planned to explore the differences between fixed‐effect and random‐effects models. However, we could not do this as the random‐effects analysis was not applicable due to a small number of included studies per each outcome. 

We also planned to use imputation methods in the sensitivity analysis to check for possible bias from missing data. Out of four included studies, only Volicer 1997 did not use the ITT principle, thus the sensitivity analysis was possible only for meta‐analysis of change in weight.

Summary of findings and assessment of the certainty of the evidence

We used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach to assess the quality of evidence behind each of our effect estimates (high, moderate, low, or very low quality) (Guyatt 2011). This rating indicates our confidence that the effect estimate is close to the true effect size, taking account of risk of bias, inconsistency between studies, imprecision in the effect estimate, indirectness with respect to our outcome of interest, and publication bias.

Key findings for each comparison in this systematic review are summarised in the ‘Summary of findings’ table, which presents a maximum of seven key results (Schünemann 2011).

The following outcomes are included in the ‘Summary of findings' table.

1. Cognitive function.

2. Behavioural and psychological symptoms.

3. Adverse events ‐ specifically, adverse events categorised as nervous system disorders, psychiatric disorders, gastrointestinal disorders, and general adverse events.

4. Agitatated or aggressive behaviours.

For each outcome in the ‘Summary of findings' table, we included the estimate of the treatment effect, the quantity of supporting evidence, and the quality of that evidence assessed via the GRADE approach (Guyatt 2011).

DBK and LP prepared a ’Summary of findings' table for pre‐defined primary outcomes. To create the 'Summary of findings' table, we used the GRADEpro GDT web application (GRADEpro GDT 2020).

Results

Description of studies

Results of the search

Our searches, conducted by the Cochrane Dementia and Cognitive Improvement Group information specialists, yielded a total of 2940 records. After de‐duplication, there were 2448 records to assess based on title and abstract screening. From this, we identified 55 references for which we sought further information. We obtained the full texts when possible and identified four studies eligible for inclusion (Herrmann 2019; van den Elsen NCT01302340; van den Elsen NCT01608217; Volicer 1997), seven ongoing studies ( ACTRN12619000474156 2019; Euctr  2010‐024577‐39 2011; Euctr 2019‐002106‐52‐GB; Euctr 2020‐001056‐17‐GB; Forester 2017; NCT04436081 2020; NCT04516057 2020) and one study that had been completed but for which for the results are not yet available (NCT03328676 2017). We excluded 28 studies. See Figure 1 for the study flow diagram and Appendix 1 for additional details of the searches we performed.

1.

Study flow diagram

Included studies

The characteristics of the four studies included in this review are summarised in the Characteristics of included studies table.

We included four small studies with a total of 126 randomised participants (Herrmann 2019; van den Elsen NCT01302340; van den Elsen NCT01608217;  Volicer 1997). Three studies were cross‐over RCTs (Herrmann 2019;  van den Elsen NCT01302340; Volicer 1997), and one was a parallel‐group trial (van den Elsen NCT01608217). All studies compared cannabinoids with placebo. Analysed interventions were the Namisol tablet, which contains pure, natural 98% delta‐THC in fixed dosages (Echo Pharmaceuticals B.V.) (van den Elsen NCT01302340; van den Elsen NCT01608217), the synthetic oral THC analogue nabilone (Herrmann 2019), and the synthetic oral THC analogue dronabinol (Marionol, Roxane Laboratories) (Volicer 1997). 

Most participants had Alzheimer's disease, and few had vascular dementia or mixed dementia.

Two studies reported non‐commercial funding (Herrmann 2019; van den Elsen NCT01302340), and two studies were conducted with the support of both commercial and non‐commercial funding (van den Elsen NCT01608217; Volicer 1997).

The oldest study was conducted in the USA (Volicer 1997). It enrolled 15 participants hospitalised in the Dementia Study Unit. Participants were diagnosed with AD, with duration of illness varying from 2 to 16 years, and, in most cases, with severe dementia. The study used a cross‐over design, with each treatment period lasting six weeks (total trial duration: 12 weeks). Participants received dronabinol or placebo on a fixed‐dose schedule, 2.5‐mg capsule or identical‐looking placebo capsule, every morning and noon. 

van den Elsen NCT01302340 was conducted at two institutions in The Netherlands. It enrolled 22 participants, but recruitment details were not provided. After the first 10 patients were admitted to the hospital, the study protocol was revised, replacing admissions to the "ambulatory group," including a five‐hour day clinic visit, a follow‐up phone call, and a home visit. This randomised, double‐blind, placebo‐controlled, repeated cross‐over trial consisted of six treatment blocks of two weeks each. Within each block, Namisol (0.75 mg twice daily in blocks one to three, and 1.5 mg twice daily in blocks four to six) and placebo were administered in random order for three consecutive days, followed by a four‐day washout. Among participants, 18 had AD, 1 had vascular dementia, and 1 had mixed dementia.

van den Elsen NCT01608217 was conducted in The Netherlands. Participants were recruited from nine participating institutes throughout the southeast of The Netherlands, including geriatric outpatient clinics, psychiatric clinics, nursing homes (including in total six locations), and a regional network of integrated care for community‐dwelling patients with dementia. A total of 24 participants were allocated in the Namisol group and 26 in the placebo group. Participants received Namisol 1.5 mg or matched placebo (1:1) three times daily for three weeks (total daily oral dose: 4.5 mg). Among participants, 34 had AD, 7 had vascular dementia, and 9 had mixed dementia.

Herrmann 2019 was conducted in Canada. It enrolled 39 participants with moderate to severe AD. Participants were recruited from a long‐term care facility and geriatric psychiatry clinics into a 14‐week randomized double‐blind cross‐over trial comparing nabilone to placebo (six weeks each) with a one‐week washout between phases. Therapeutic dose was 1 to 2 mg per day. In the treatment phase, after placebo run‐in, participants received 0.25 mg before bedtime for three nights, then were titrated up to 0.5 mg/d total for the next four days, and maintained this dose for Week 1. In Week 2, participants were titrated up to 1 mg/d. During Weeks 3 and 4, participants received a flexible dose of 1 to 2 mg/d based on tolerability.

Excluded studies

We excluded 28 studies as described in the Characteristics of excluded studies table. These studies were excluded because they were not RCTs (Amanullah 2013; Assogna 2020; Boxer 2015; Broers 2019; Ghaffar 2008; Kahraman 2009; Libro 2016; Noonan 2010; Passmore 2008; Scotter 2010; Shelef 2016; Walther 2006; Walther 2011; Wilhelm 2017; Woodward 2014; Zajac 2015), they did not include participants with dementia (Aragona 2009; Caldentey 2012; Chagas 2013; Chagas 2014; Consroe 1991; Curtis 2009; NCT00842985 2008; NCT01964547 2012; Solowij 2014; Stone 2010), or they were terminated prematurely (Euctr 2016). One study was excluded because randomisation was not reported, and results for the cannabinoid group were not reported separately from those for another intervention group (Mahlberg 2007).

Risk of bias in included studies

We summarised the 'Risk of bias' results in Figure 2 and Figure 3and Figure 3. For detailed assessment, please see the Characteristics of included studies table.

2.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

3.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

Random sequence generation: all studies were described as randomised; three studies reported adequate methods of randomisation and were judged to have low risk of bias (Herrmann 2019; van den Elsen NCT01302340; van den Elsen NCT01608217); one study did not report a method of randomisation and thus was judged to have unclear risk of bias (Volicer 1997).

Allocation concealment: three studies used an adequate allocation concealment method and were judged to have low risk of bias (Herrmann 2019; van den Elsen NCT01302340; van den Elsen NCT01608217); the remaining study did not report method of allocation concealment and thus was judged to have unclear risk of bias for this domain (Volicer 1997).

Blinding

All studies were described as double‐blind. For three studies, we obtained detailed information about which groups of individuals were blinded (Herrmann 2019;van den Elsen NCT01302340; van den Elsen NCT01608217); thus we judged these studies as having low risk of bias. The remaining study reported only that the study was double‐blind (Volicer 1997), but this term is used inconsistently and makes it difficult to know who was blinded; we judged this study to have unclear risk of bias.

Incomplete outcome data

Three studies had low attrition (Herrmann 2019; van den Elsen NCT01302340; van den Elsen NCT01608217); thus we judged them to have low risk of bias. In one study (Volicer 1997), of 15 patients enrolled, 11 patients completed both study periods. One patient who died of a heart attack two weeks before the end of the study while on placebo was also included in the analysis. It is not reported when participants who dropped out due to intercurrent infection terminated the study. We judged this study to have unclear risk of bias for this domain.

Selective reporting

Three studies had reported that the study protocol was published; we did not find differences between outcomes planned in those study protocols and outcomes reported in manuscripts (Herrmann 2019; van den Elsen NCT01302340; van den Elsen NCT01608217); thus we judged them to have low risk of bias for selective reporting. Volicer 1997 did not report information about availability of the study protocol; thus we were unable to check whether there was selective reporting; we judged this study to have unclear risk of bias for selective reporting.

Other potential sources of bias

We did not find other potential sources of bias. Thus, all studies were judged as having low risk of other bias.

Effects of interventions

See: Table 1

See Table 1.

Primary outcomes

1. Changes in global and specific cognitive functions

Two studies reported changes in global or specific cognitive functions (Herrmann 2019; van den Elsen NCT01608217). van den Elsen NCT01608217 assessed episodic memory scores in dementia patients using the Paired Associate Learning Wechsler Memory Scale‐Revised. However, as scores were available for only a minority of enrolled patients (18% or 36%) and information was insufficient to extract their variability, this study was not included in quantitative analysis for this outcome. Herrmann 2019 used the Standardized Mini‐Mental State Examination (sMMSE) to assess global cognitive function in a cross‐over study comparing nabilone to placebo. This study found a mean difference (MD) between groups of 1.1 points on the sMMSE scale in favour of nabilone (95% confidence interval (CI) 0.1 to 2.1; 1 cross‐over trial, 38 participants; Analysis 1.1). A minimum clinically important difference (MCID) of 1.4 sMMSE points has been suggested for dementia patients (Howard 2011), although this MCID was estimated in a single RCT assessing patients at a particular stage of dementia. However, the difference of 1.1 on the 30‐point sMMSE scale is unlikely to be clinically important. We judged the certainty of evidence to be very low due to indirectness and imprecision.

1.1. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 1: Changes in global and specific cognitive function, sMMSE scale

2. Overall behavioural and psychological symptoms of dementia

Three studies reported overall behavioral and psychological symptoms of dementia (Herrmann 2019; van den Elsen NCT01302340; van den Elsen NCT01608217). Two of the studies used NPI total score to assess dementia symptoms, whereas Herrmann 2019 used a modified version of the scale ‐ NPI‐NH total scores. Both versions of the NPI questionnaire are based on the same 12 domains, with the key difference that the NPI‐NH is assessed by nursing home staff instead of family caregivers/researchers. The pooled MD estimate demonstrated there may be a very small difference in overall behavioural and psychological symptoms of dementia in favour of cannabinoids (‐1.97, 95% CI ‐3.87 to ‐0.07; 1 parallel‐group and 2 cross‐over studies, 110 participants; Analysis 1.2). We judged the certainty of evidence to be low due to imprecision and inconsistency. We noted substantial heterogeneity between studies (P = 0.04, I² = 69%), with potential sources of heterogeneity including the type of questionnaire used, the severity of dementia, and the type of cannabinoid. Although symptom domains covered by the NPI and NPI‐NH scales are identical, questions have been rephrased to allow an assessment of the impact of behavioural disturbances on professional caregivers. These changes, together with the predominant type of assessor (family/professional caregiver), might affect the effect size. In addition, there was a difference in severity of dementia symptoms between studies, with Herrmann 2019 including on average more severely ill patients (mean sMMSE ± SD of 6.5 ± 6.8) than the other two studies (mean MMSE ± SD at baseline of 16.9 ± 7.8  in van den Elsen NCT01302340 and 14.8 ± 6.7 in van den Elsen NCT01608217). Finally, although Herrmann 2019 used the synthetic oral THC analog Nabilone, van den Elsen NCT01302340 and van den Elsen NCT01608217 used the oral THC Namisol.

1.2. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 2: Overall behavioural and psychological symptoms of dementia, NPI  and NPI‐NH total score

We conducted a subgroup analysis considering Herrmann 2019 as a separate subgroup because of differences in participants, interventions, and outcome measures (Analysis 1.3). However, given the very small number of studies and the range of differences between them, it is not possible to draw conclusions about the origin of the heterogeneity. The pooled effect in the two studies that included less severely ill patients and assessed the effect of oral THC using the NPI was an MD of 0.01 points (95% CI ‐2.51 to 2.52; P = 0.32, I² = 0%; 1 parallel and 1 cross‐over study, 72 participants). In Herrmann 2019 (more severely ill patients, synthetic analogue of THC, outcome assessed with NPI‐NH), study authors reported there may be a small improvement in symptoms among severely ill patients receiving Nabilone compared to those given placebo (MD ‐4.6 points, 95% CI ‐7.5 to ‐1.7; 1 cross‐over trial, 38 participants). An MCID of 8 points on the NPI has been suggested (Howard 2011). Again, this MCID is estimated in a single study and should not be taken as firmly established, but it is likely that neither the effect in the pooled analysis nor effects in either subgroup are clinically important.

1.3. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 3: Subgroup analysis ‐ Overall behavioural and psychological symptoms of dementia, NPI  and NPI‐NH total score

3. Adverse events

All included studies reported data on adverse events. However, the total number of severe adverse events, the total number of mild and moderate adverse events, and the total number of serious adverse events (SAEs) were not reported in a way that permitted meta‐analysis. As we were most often unable to use data from cross‐over trials to analyse a summary effect of cannabinoids on specific types of adverse events ‐ nervous system disorders, psychiatric disorders, gastrointestinal disorders, and general symptoms ‐ we showed the effect size and the 95% CI using data from the only parallel‐group trial (van den Elsen NCT01608217). Only for sedation (including lethargy) and treatment‐limiting sedation were we able to use data from the cross‐over trial (Herrmann 2019).

3. 1. Total number of adverse events 

Volicer 1997, which included 12 participants in a cross‐over study, reported 67 adverse events among participants taking dronabinol and 58 adverse events among those given placebo;  study authors did not report adverse events for separate study periods.

van den Elsen NCT01302340, which included 22 participants in a cross‐over study, reported 46 adverse events among participants taking Namisol and 48 adverse events among participants taking placebo during Period A (the first period of six weeks). There were 45 adverse events with Namisol and 45 adverse events with placebo during Period B (second period of six weeks). 

van den Elsen NCT01608217 reported 16 adverse events in the Namisol group (N = 24) and 14 adverse events in the placebo group (N = 26).

Herrmann 2019, which included 39 patients in a cross‐over study, reported treatment‐emergent adverse events (TEAEs) in 38 patients, as 1 patient discontinued the study during the placebo run‐in (Week 1) due to clinically significant delusions and was not included in the analysis; there were 31 TEAEs with nabilone and 14 TEAEs with placebo; the study did not report TEAEs for different study periods. Study authors reported the results of McNemar's test with P = 0.05.

The certainty of evidence for the total number of adverse events was judged as very low ‐ downgraded by three levels due to very serious concerns about indirectness (short duration of studies) and serious concerns about imprecision (multiple small trials).

3.2. Mild to moderate adverse events

Volicer 1997, which enrolled 15 participants in a cross‐over study, reported 66 mild to moderate adverse events among participants taking dronabinol and 58 among those given placebo; study authors did not report these data for separate study periods.

van den Elsen NCT01302340 and van den Elsen NCT01608217 reported exactly the same numbers of mild to moderate adverse events with Namisol and placebo, as for total adverse events. van den Elsen NCT01302340 additionally reported an incidence rate ratio (IRR) for mild to moderate adverse events of 0.96 (95% CI 0.7 to 1.3), whereas van den Elsen NCT01608217 compared numbers of patients with at least one unique mild or moderate episode using the Chi² test (P = 0.36). Due to different units of observation and use of the Chi² test on paired data, we did not extract and meta‐analyse these data.  

Herrmann 2019 did not report adverse events as "mild to moderate."

We judged the certainty of evidence for mild to moderate adverse events as very low ‐ downgraded by three levels due to very serious concerns about indirectness (short duration of studies) and serious concerns about imprecision (multiple small trials).

3.3. Serious adverse events

Volicer 1997, which enrolled 15 participants in a cross‐over study, reported that 1 patient died of a heart attack 2 weeks before the end of the study while on placebo. This study was terminated for 3 patients: 1 developed a grand mal seizure after the first dronabinol dose, and 2 developed a serious intercurrent infection.

van den Elsen NCT01302340 and van den Elsen NCT01608217 reported there were no serious adverse events.

Herrmann 2019 (38 participants in a cross‐over study) reported five severe adverse events with nabilone (lethargy in 2 patients, death in 1 patient, critically high INR in 1 patient, and myocardial infarction in 1 patient) and four with placebo (death in 1 patient, cancer diagnosis in 1 patient, pneumothorax in 1 patient, and sepsis due to UTI in 1 patient). Study authors reported the exact P value of McNemar's test as 0.69. The frequency distribution of SAEs per individual and period that corresponds to this P value indicates that 1 person experienced an SAE during both nabilone and placebo periods. As such observation contradicts the definition of SAE (participation in the study should be stopped after experiencing it), we decided we would not report this effect size.  

We judged the certainty of evidence for SAEs as very low ‐ downgraded by three levels due to very serious concerns about indirectness (short duration of studies) and serious concerns about imprecision (multiple small trials).

3.4. Adverse events ‐ nervous system disorders

The parallel‐group study van den Elsen NCT01608217 reported the total number of nervous system disorders as adverse events. We considered the evidence to be of low certainty (downgraded by two levels for imprecision: one small trial, and confidence interval crosses the line of no effect). Therefore we are uncertain of the effects of Namisol on nervous system event disorders as adverse events (odds ratio (OR) 0.71, 95% CI 0.23 to 2.18;  P = 0.59; 1 parallel study, 50 participants; Analysis 1.4). 

1.4. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 4: Adverse events ‐ nervous system disorders

The cross‐over study Herrmann 2019 (N = 38) reported dizziness in one patient receiving nabilone and none in those receiving placebo.

Evidence from van den Elsen NCT01608217 was considered to be of low certainty (downgraded by two levels for imprecision: one small trial and confidence interval crosses the line of no effect). Therefore we are uncertain of the effect of Namisol on the rate of somnolence (OR 0.50, 95% CI 0.08 to 3.02; P = 0.45; 1 parallel study, 50 participants; Analysis 1.4). In the Volicer 1997 cross‐over study (N = 12), somnolence was experienced by eight participants while taking dronabinol and by four participants given placebo.

Herrmann 2019 reported the number of patients experiencing sedation (including lethargy) while taking nabilone (N = 17) compared to placebo (N = 6) (OR 2.83, 95% CI 1.07 to 7.48; 1 cross‐over study, 38 participants; Analysis 1.5). The results, which were imprecise due to a small study sample, showed that nabilone intervention increases the odds of sedation by 183%, with the 95% CI allowing for a small 7% to a large and clinically important increase of odds by 648% during nabilone treatment. When only treatment‐limiting sedation was analysed, study authors observed five such events during nabilone and one during placebo treatments, but likely due to a small sample size, the confidence interval was very large, crossing the line of no effect (OR 4.01, 95% CI 0.40 to 40.56; Analysis 1.6); this imprecision affects our certainty of the evidence; further information would be needed before we could draw a more certain conclusion.

1.5. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 5: Adverse events ‐ Sedation/lethargy

1.6. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 6: Adverse events ‐ treatment‐induced sedation

The rate of muscle spasms, as reported in van den Elsen NCT01608217 (OR 0.35, 95% CI 0.01 to 8.93; P = 0.52; 1 parallel study, 50 participants; Analysis 1.4), was considered to be of low certainty (downgraded by two levels for imprecision: one small trial and confidence interval crosses the line of no effect).

3.5. Adverse events ‐ psychiatric disorders

The parallel‐group study van den Elsen NCT01608217 reported the total number of participants experiencing psychiatric disorders as adverse events (OR 2.26, 95% CI 0.57 to 9.02; 1 parallel study, 50 participants; Analysis 1.7). We judged the certainty of the evidence as low (downgraded by two levels for imprecision: one small trial and confidence interval crosses the line of no effect); thus, we are uncertain of this result.

1.7. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 7: Adverse events ‐ psychiatric disorders

The rate of euphoria in van den Elsen NCT01608217 was not different between participants taking Namisol or placebo (OR 0.35, 95% CI 0.01 to 8.93; P = 0.70; 1 parallel study, 502 participants; Analysis 1.7). The cross‐over study of Volicer 1997 reported euphoria in seven participants taking dronabinol and in five participants taking placebo (out of 12 patients total). We judged the certainty of evidence as low (downgraded by two levels for imprecision: one small trial and confidence interval crosses the line of no effect); thus, we are uncertain of this result.

3.6. Adverse events ‐ gastrointestinal disorders

van den Elsen NCT01608217 reported total numbers of participants experiencing gastrointestinal disorders as adverse events between the group taking Namisol and the group taking placebo (OR 2.40, 95% CI 0.40 to 14.49; P = 0.34; 1 parallel study, 50 participants; Analysis 1.8). We judged the certainty of evidence as low (downgraded by two levels for imprecision: one small trial and confidence interval crosses the line of no effect); thus, we are uncertain of this result.

1.8. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 8: Adverse events ‐ gastrointestinal disorders

van den Elsen NCT01608217 also reported total numbers of participants experiencing nausea between the group taking Namisol and the group taking placebo (OR 2.27, 95% CI 0.19 to 26.81; P = 0.51; 1 parallel study, 50 participants; Analysis 1.8). We judged the certainty of evidence as low (downgraded by two levels for imprecision: one small trial and confidence interval crosses the line of no effect); thus, we are uncertain of this result.

3.7. Adverse events ‐ other

van den Elsen NCT01608217 reported total numbers of participants experiencing other adverse events between the group taking Namisol and the group given placebo (OR 0.70, 95% CI 0.11 to 4.58; P = 0.71; 1 parallel study, 50 participants; Analysis 1.9). We judged the certainty of evidence as low (downgraded by two levels for imprecision: one small trial and confidence interval crosses the line of no effect); thus, we are uncertain of this result.

1.9. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 9: Adverse events ‐ other

van den Elsen NCT01608217 did not find a difference in the number of participants experiencing fatigue between groups taking Namisol and those given placebo (OR 0.70, 95% CI 0.11 to 4.58; P = 0.71; 1 parallel study, 50 participants; Analysis 1.9). Out of 12 patients in the cross‐over study of Volicer 1997, 9 participants taking dronabinol and 5 participants taking placebo experienced tiredness. We judged the certainty of evidence as low (downgraded by two levels for imprecision: one small trial and confidence interval crosses the line of no effect); thus, we are uncertain of this result.

Secondary outcomes

4. Changes in functional outcomes

van den Elsen NCT01608217 reported effects of Namisol on changes in functional outcomes using the Barthel Index. This study demonstrated on average little to no effect of Namisol on improving functional outcomes (MD 0.6, 95% CI ‐0.75 to 1.95; 1 parallel‐group study, 50 participants; Analysis 1.10). We are uncertain about the effects of cannabinoids on functional outcomes (low certainty; downgradeded by two levels due to imprecision because of one small study and confidence interval crossing the line of no effect).

1.10. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 10: Changes in functional outcomes, Barthel Index

5. Overall dementia severity

Two studies reported the percentage of patients with minimal to marked improvement in dementia severity from baseline (Herrmann 2019; van den Elsen NCT01608217), as assessed by the Clinician’s Global Impression of Change (CGIC) scale. Pooled results (OR 1.88, 95% CI 1.03 to 3.44; P = 0.02, I² = 81%; 2 studies ‐ 1 parallel and 1 cross‐over, 89 participants) suggest that odds of cannabinoids improving dementia status increased by 88%, with effects possibly ranging from little, clinically non‐important effects to large, clinically important ones (Analysis 1.11). The estimate was of low certainty, with both studies having wide 95% CIs and with considerable statistical heterogeneity among studies. Thus, we are uncertain about the extent of the effect of cannabinoids on overall dementia severity.

1.11. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 11: Minimal to marked improvement on dementia severity, assessed on the CGI Change scale

6. Objective sleep outcomes measured with polysomnography or actigraphy

None of the included studies reported data for this outcome.

7. Change in appetite

Change in appetite domain was evaluated by four outcomes: weight, nutritional status assessed with the Mini‐Nutritional Assessment Short‐Form scale (MNA‐SF), body mass index (BMI), and caloric intake.

Three studies reported differences in weight (Herrmann 2019; van den Elsen NCT01608217; Volicer 1997). Overall, small to no difference in weight was found with cannabinoids compared to placebo (MD 0.33 kg, 95% CI ‐0.08 to 0.75; P = 0.04, I² = 68%; 3 studies ‐ 1 parallel and 2 cross‐over, 100 participants; Analysis 1.12). The observed difference was within the range that excluded any clinically important effect on weight. After the sensitivity analysis was applied for missing data by excluding Volicer 1997, which was the only study that did not perform ITT analysis, we observed a comparable summary effect (MD ‐0.04, 95% CI ‐0.55 to 0.47; P = 0.83, I² = 0%; 2 studies ‐ 1 parallel and 1 cross‐over, 88 participants) supporting the finding of no effect of cannabinoids on weight.  We are uncertain about the effects of cannabinoids on appetite (low certainty; downgraded by two levels due to imprecision because of small studies and confidence interval crossing the line of no effect).

1.12. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 12: Weight [kg]

Differences in effects of cannabinoids and placebo on nutritional status, estimated with the MNA‐SF, and on BMI were assessed only in one cross‐over study (Herrmann 2019). This study demonstrated an increase in nutritional status (MD 0.2, 95% CI 0.02 to 0.38; 1 cross‐over study, 38 participants), but again the range of observed differences, with the maximum of 2.7% change on the MNA‐SF (Analysis 1.13). We considered that this was not a clinically important effect. Data from a single study affect our certainty about the result (moderate‐certainty evidence, downgraded by one level due to imprecision). Cannabinoids likely result in little to no difference regarding nutritional status.

1.13. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 13: MNA‐SF

Data from a single study affect our certainty about the result (moderate‐certainty evidence, downgraded by one level due to imprecision).

Regarding BMI,  the study reported little to no effect among participants receiving nabilone (MD ‐0.14, 95% CI ‐0.35 to 0.07; 1 cross‐over study, 38 participants; Analysis 1.14). We considered that this was not a clinically important effect. We considered the results to be of low certainty due to imprecision (downgraded by two levels for imprecision because of one small study and confidence interval crossing the line of no effect).

1.14. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 14: BMI

For the difference in caloric intake, which was assessed in Volicer 1997, it is uncertain whether there was a difference between dronabinol and placebo (19 kcal, 95% CI ‐509 to 547; 1 cross‐over study, 12 participants; Analysis 1.15). However, the only study that did not apply the ITT principle in analysis was largely underpowered to confidently detect or exclude a small or large clinically important effect, or no effect. We considered the results of low certainty due to imprecision (downgraded by two levels for imprecision because of one small study and a large confidence interval crossing the line of no effect).

1.15. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 15: Caloric intake [kcal]

8. Agitated or aggressive behaviours

Three studies (Herrmann 2019; van den Elsen NCT01302340; van den Elsen NCT01608217) analysed the effects of cannabinoids on agitated or aggressive behaviours in dementia patients, measured on the Cohen‐Mansfield Agitation Inventory (CMAI) scale (range 29 to 203). The pooled MD effect showed that cannabinoids, compared to placebo, reduced points on the CMAI scale (MD ‐2.35 points, 95% CI ‐4.10 to ‐0.60; P = 0.07, I² = 62%; 3 studies ‐ 1 parallel and 2 cross‐over, 100 participants) with substantial heterogeneity (Analysis 1.16). This reduction was within the range of clinically non‐important effects, excluding clinically relevant effects of point differences of 8 or more (Zuidema 2011). The estimate was of low certainty, with both studies having wide 95% CIs, and with considerable statistical heterogeneity among studies. Thus, we are uncertain about the extent of the effects of cannabinoids on agitated or aggressive behaviours.

1.16. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 16: CMAI 

The same three studies (Herrmann 2019; van den Elsen NCT01302340; van den Elsen NCT01608217) reported differences in agitated or aggressive behaviours using another measurement scale: NPI subscale agitation/aggression, with a range of 0 to 12 points. The pooled MD showed a reduction in points on the NPI agitation/aggression subscale among patients receiving cannabinoids compared to those given placebo (MD ‐0.63 points, 95% CI ‐1.08 to ‐0.18; P = 0.06, I² = 65%; 3 studies ‐ 1 parallel and 2 cross‐over, 100 participants), with substantial heterogeneity (Analysis 1.17). However, on this measurement scale, one could not exclude that cannabinoids may have little clinically important effect or no effect. We judged the certainty of evidence as moderate (downgraded by one level for imprecision due to small studies). Cannabinoids likely result in little to no difference in this outcome.

1.17. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 17: NPI subscale agitation/aggression

9. Mood, measured with any validated tool

Except for one study that used the Lawton Observed Affect Scale to evaluate mood in patients after receiving dronabinol or placebo (Volicer 1997), no other study assessed this domain. In Volicer 1997, the observed decrease in negative affect was higher during dronabinol treatment ‐ even more for participants who received dronabinol first. Positive affect remained similar during the 12‐week trial for both dronabinol and placebo and for both treatment periods. However, data were not adequately reported so that variability of scores could be estimated, and data were not used for quantitative analysis. We estimated certainty of these results as moderate (downgraded by one level due to imprecision ‐ one small study); dronabinol probably decreases negative affect.

10. Carer ratings of patient sleep using sleep diaries or validated observer scales

None of the included studies reported data for this outcome.

11. Quality of life

van den Elsen NCT01608217 was the only study investigating differences in quality of life. The outcome was assessed among participants after administration of Namisol or placebo, using the QoL‐AD scale (scale's range 13 to 52 points). It is uncertain whether there was improvement in quality of life among patients receiving placebo (MD ‐0.50 points, 95% CI ‐2.60 to 1.60; 1 parallel‐design study, 50 participants) with the expected magnitude of this effect in the typical patient ranging from no effect to small clinically important harm/benefit due to Namisol (Analysis 1.18). We considered the results of low certainty (downgraded by two levels for imprecision because of one small study and confidence interval crossing the line of no effect).

1.18. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 18: QoL‐AD

12. Any other symptoms associated with dementia (e.g. alterations in circadian rhythm)

None of the included studies reported data for this outcome.

13. Caregiver burden and caregiver quality of life

Two studies analysed the effects of administering cannabinoid agents to dementia patients on caregiver burden using different measurement scales. van den Elsen NCT01302340 used the Zarit Burden Interview (ZBI), whereas Herrmann 2019 used the NPI‐NH Caregiver Distress scale. Thus, we calculated the standardised mean difference. Overall, we observed small improvement in scores on caregiver burden after patients were administrated cannabinoids (standardised mean difference (SMD) ‐0.12, 95% CI ‐0.38 to 0.13; P = 0.17, I² = 48%; 2 cross‐over studies, 60 participants), suggesting that cannabinoids may have little to no clinically important effect (Analysis 1.19). We considered the results of low certainty (downgraded by two levels for imprecision because of one small study and confidence interval crossing the line of no effect).

1.19. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 19: Caregiver burden

14. Treatment or research discontinuation/dropout (as measures of acceptability)

Three studies reported data on all‐cause discontinuation by a treatment group (Herrmann 2019; van den Elsen NCT01302340; van den Elsen NCT01608217). van den Elsen NCT01302340, which reported zero discontinuation in both arms, was not included in the analysis. The other two studies were substantially underpowered for this outcome. This resulted in a largely uncertain pooled estimate (OR 1.02, 95% CI 0.33 to 3.13; P = 0.54, I² = 0; 2 studies ‐ 1 parallel and 1 cross‐over study, 88 participants), indicating that at this point, we do not know whether cannabinoids have any effect on discontinuation because both clinically important effects of cannabinoids and no effect could be expected (Analysis 1.20). We considered the results of low certainty (downgraded by two levels for imprecision because of one small study and confidence interval crossing the line of no effect).

1.20. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 20: All‐cause discontinuation

15. Mortality

Four studies reported on all‐cause mortality (Herrmann 2019;  van den Elsen NCT01302340; van den Elsen NCT01608217; Volicer 1997), but only two had non‐zero events in both arms and were included in the meta‐analysis (Herrmann 2019; Volicer 1997). Because these studies were substantially underpowered for mortality, the resulting pooled estimate was uncertain (OR 0.59, 95% CI 0.07 to 4.62; P = 0.59, I² = 0; 2 cross‐over studies, 50 participants; Analysis 1.21). We considered the results of low certainty (downgraded by two levels for imprecision because of one small study and wide confidence interval crossing the line of no effect).

1.21. Analysis.

Comparison 1: Comparison: cannabinoids versus placebo, Outcome 21: All‐cause mortality

Discussion

Summary of main results

We included four small studies with a total of 126 randomised participants (Herrmann 2019;  van den Elsen NCT01302340; van den Elsen NCT01608217; Volicer 1997.

Based on our summary data, it is uncertain whether cannabinoids can have beneficial or harmful effects for people with dementia, compared to placebo. For few outcomes suggesting a positive effect of cannabinoids, the effect possibly ranged from a small effect that is not clinically meaningful to a large, clinically important one. Included studies were underpowered, heterogeneity among them was considerable, and their results were inconsistent.

Although there was no significant difference between cannabinoids and total numbers of nervous system, psychiatric and gastrointestinal adverse events, it should be noted that one cross‐over trial reported a significant difference in experiencing sedation with nabilone compared to placebo.

The summary effect sizes reported were often smaller than what is considered the minimum clinically important difference for the relevant scale. This may be due in part to relatively short follow‐up across studies.

Because the included studies used different types of cannabinoids, different dosage and frequency, and different duration of administration of cannabinoids, and because they did not report data for relevant subgroups (e.g. different types of dementia in studies that included multiple types of dementia), we were unable to perform pre‐planned subgroup analyses regarding dose of the intervention, type of dementia, and stage of dementia because of the small number of studies and potential sources of heterogeneity. 

Overall completeness and applicability of evidence

For most of our pre‐specified outcomes of interest, we found data in at least one of the included trials, but the studies were very small and results uncertain. For example, only one study reported data on patients' quality of life, and two studies reported data on caregiver burden. Furthermore, none of the trials reported data for some of our pre‐specified outcomes, including sleep outcomes, carer ratings of patients' sleep, caregiver quality of life, or any other symptoms associated with dementia (e.g. alterations in circadian rhythm). Reporting data for adverse events was inconsistent, and we could not always use them in summary analyses. Thus, the data identified from the four included trials cannot be considered complete for relevant outcomes.

Most of the included participants had Alzheimer's disease (AD). Thus, the results of this review are not necessarily applicable to individuals living with other types of dementia.

Although some of the included participants had mild dementia and were recruited from a community setting, most participants had moderate to severe dementia, which often requires institutionalisation. Thus, study findings may not apply to all individuals living with dementia.

One study was conducted in Canada, one in the USA, and two in The Netherlands. Thus, the evidence may not apply to other settings. Dementia care is different across international settings, and some of the treatments used routinely in the included studies are not recommended or are not available in all healthcare systems.

Three of the included studies were more recent ‐ conducted over the past decade; Herrmann 2019 enrolled participants from 2015 to 2017, van den Elsen NCT01302340 enrolled participants from 2011 to 2013, and van den Elsen NCT01608217 enrolled participants from 2021 to 2014. The fourth study, Volicer 1997, did not report enrolment dates but was published more than 20 years ago. Best practice in dementia care is changing over time, and the generalisability of studies that are decades old is questionable.

All interventions in the included studies were oral cannabinoid medicines; thus, results of this review may not apply to other routes of cannabinoid administration. Furthermore, all medicines in the included trials contained delta‐THC, natural or synthetic; thus the review results may not apply to other types of active cannabinoids such as cannabidiol.

Quality of the evidence

This review included four trials ‐ three cross‐over and one parallel‐group trial. All four trials were placebo‐controlled randomised controlled trials (RCTs), and sample sizes in all four included trials were very small, ranging from 15 to 50 enrolled participants. Thus, we have a lot of uncertainty about their results. Using GRADE methods, we judged the certainty of evidence for primary outcomes to be low or very low due to risk of bias, inconsistency, indirectness, and imprecision of results.

Potential biases in the review process

We adhered to the recommended Cochrane search and review methods; thus we consider that we have identified all data of interest to the review question. 

We are unable to exclude the possibility of publication bias.

Agreements and disagreements with other studies or reviews

Multiple reviews have been published recently on the effects of cannabinoids for dementia (Bahji 2020; Charenbron 2021; Hillen 2019; Lim 2017; Liu 2015; Liu 2016; Peprah & McCormack 2019; Ruthirakuhan 2019; Sherman 2018; Walther & Halpern 2010; Weier et Hall 2017), and our review generally agrees with them, sharing the view that available evidence is inconclusive and insufficient, and that more RCTs on this topic are needed.

In 2010, Walther and Halpern published a review of clinical and preclinical data on cannabinoid use in different neurodegenerative diseases, including AD.  Despite the possible therapeutical potential of cannabinoids in the in vitro studies, review authors found scarce clinical data for the use of cannabinoids in AD. Clinical studies focused mainly on symptomatic treatment of, for example, behavioural and motor symptoms, showing somewhat promising results in one small RCT (Volicer 1997), which was also included in our review; one open‐label study (Walther 2006); and one case report (Passmore 2008). Review authors concluded there is a need for further research in larger clinical trials with longer follow‐up periods (Walther & Halpern 2010).

A review of Liu et al. from 2015 assessed the effects of cannabinoids on agitation and aggression in AD. Review authors included letters, case studies, and controlled trials from four electronic databases. Although findings from the six included studies showed significant benefit from synthetic cannabinoids ‐ dronabinol or nabilone ‐ for agitation and aggression, review authors concluded that definitive conclusions were limited by small sample sizes, short trial duration, and lack of placebo control in some studies (Liu 2015). Out of these six studies, five were excluded from our review: three were not RCTs (Passmore 2008; Walther 2006; Woodward 2014), one was a prematurely finished RCT with 2 participants (Walther 2011), and for one, randomisation was not reported and results were not reported separately for the dronabinol group (Mahlberg 2007).

Current and novel treatments for pharmacological management of agitation and aggression in AD included cannabinoids as one group of those promising novel treatments reviewed by Liu et al. in 2016. Out of the six studies mentioned in the review,  we excluded four (Passmore 2008; Walther 2006; Walther 2011; Woodward 2014). The review highlighted that although no large RCTs had tested cannabinoids for treating agitation in AD, the results of available individual case reports and small studies suggested that cannabinoids may be a potentially useful intervention, acting on an alternative pathway in which to target symptoms, such as agitation and aggression, associated with AD (Liu 2016).

In 2017, Lim et al. published a systematic review of the effectiveness of medical cannabis for the treatment of different psychiatric, movement, and neurodegenerative disorders. This review included four RCTs investigating cannabinoids (dronabinol, Namisol) in dementia, one of which we excluded as prematurely terminated (Walther 2011). Despite some positive findings regarding agitation in AD and dementia,  review authors stated that a definitive conclusion could not be drawn due to low‐quality evidence associated with various methodological issues in the included trials (Lim 2017).

A 2017 review by Weier and Hall concluded that available evidence from two analysed RCTs (van den Elsen NCT01302340; van den Elsen NCT01608217) suggests that the potential of cannabinoids may be overstated. This review also included open‐label trials and case reports, and review authors found that those studies indicate that cannabinoids may provide some therapeutic benefit for individuals with dementia. We excluded these studies from our review as they were not RCTs (Amanullah 2013; Passmore 2008; Shelef 2016; Walther 2006). Review authors highlighted that common risks of long‐term cannabis use, such as cognitive decline, may not be a concern for an individual suffering from advanced dementia; however, other side effects such as psychosis are of uncertain significance for an individual with dementia. They urged that new trials in this field are needed, so that cannabinoids can be seriously considered as an alternative to current pharmacotherapies for treating symptoms of dementia (Weier et Hall 2017).

A review of the effect of cannabinoids on neuropsychiatric symptoms, pain, and weight loss in dementia was published in 2018 by Sherman et al. This review included seven studies and one case report. Of these, four were excluded, as they were not RCTs (Mahlberg 2007; Passmore 2008; Walther 2006; Woodward 2014), and one was prematurely finished (Walther 2011). Review authors found that the included studies had small sample sizes and included various clinical populations. They concluded that their findings suggest that cannabinoids may be well tolerated and effective for the treatment of neuropsychiatric symptoms such as agitation, as well as weight and pain management, for patients with dementia. However, the review authors also concluded that additional studies are needed to elucidate further the relative risks and benefits of cannabinoids for dementia (Sherman 2018). 

A meta‐analysis of Ruthirakuhan et al., published in 2019, investigated the efficacy of cannabinoids for agitation and aggression in patients with AD. Based on data from six studies, review authors concluded that their meta‐analysis provided little evidence about the efficacy of cannabinoids for agitation in AD. Their findings indicate that THC for the treatment of agitation has been consistently negative; results with synthetic cannabinoids were inconclusive due to substantial heterogeneity. Review authors warned that the safety of cannabinoids should be closely monitored, as they are associated with increased sedation (Ruthirakuhan 2019). Of these six studies, two were excluded from our review: one was a prematurely finished trial with two participants (Walther 2011), and the other did not report randomisation (Mahlberg 2007).

In the 2019 systematic review assessing the safety and effectiveness of cannabinoids for treating neuropsychiatric symptoms in dementia, Hillen et al. included 12 studies (6 RCTs, 2 cohort studies, and 4 case studies/series). Of the six included studies, we excluded one as a prematurely ended trial (Walther 2011),  and three were published manuscripts from a single RCT (van den Elsen NCT01302340). Hillen et al. reported high risk of bias in eight of their included studies. Although the only RCT with low risk of bias did not show better efficacy of cannabinoids compared to placebo (van den Elsen NCT01608217), review authors report that some of the observational studies had promising results. Taking into account the use of low doses of oral cannabinoids and the favourable safety profile with only mild adverse events reported in the included studies, review authors suggested that higher doses of cannabinoids and different drug formulations should be used in future trials (Hillen 2019). 

Peprah and McCormack conducted a review of clinical effectiveness and guidelines on medical cannabis for dementia in 2019, including results from one previous systematic review (Hillen 2019), as well as from one prospective observational pilot study (Broers 2019). With only limited and low‐quality evidence of cannabinoid effectiveness for treatment of various neuropsychiatric symptoms, review authors reported no relevant evidence‐based clinical guidelines for the use of cannabinoids for treatment of dementia (Peprah & McCormack 2019). 

In 2020, Bahji et al. published a systematic review and meta‐analysis about the effects of cannabinoids on NPS symptoms of dementia. They included 9 studies, RCTs, and quasi‐randomised studies in their review. Of the six included RCTs, we excluded one as prematurely finished (Walther 2011), along with one that reported no randomisation methods, without separate results for dronabinol and melatonin groups (Mahlberg 2007). The review authors highlighted that they identified only a few small studies; few could be meta‐analysed due to inconsistent reporting styles across studies. The overall quality of included studies was judged as low. Review authors concluded that there is preliminary evidence for the efficacy and tolerability of cannabinoids as interventions for NPS. Additionally, they advised that population‐based studies are needed to characterise the real‐world effectiveness and acceptability of cannabinoids for dementia (Bahji 2020).

In 2021, Charenbron et al. reviewed evidence from RCTs on the effectiveness of cannabinoids for dementia. Based on five studies (one excluded from our review as prematurely terminated ‐ Walther 2011), they concluded that nabilone might be useful for treating agitation in patients with dementia, but that there is no convincing evidence for THC. Review authors recommended that more studies are needed to clarify further and to better assess the benefits of cannabinoids for dementia. The review concluded that it may be too early to postulate that cannabinoids have any effect on dementia symptoms or on their progression (Charenbron 2021).

Authors' conclusions

Implications for practice.

Based on data from four small, heterogeneous, and short placebo‐controlled trials, it is uncertain whether cannabinoids have any beneficial or harmful effects on dementia compared to placebo. If there are benefits of cannabinoids for people with dementia, the effects may be too small to be clinically meaningful.

Implications for research.

Our review demonstrates the need for new, well‐designed trials to evaluate the efficacy, safety, and economics of cannabinoids for individuals with dementia.

Based on this review, we can offer some guidance for future RCTs on cannabinoids for dementia. Future studies should:

1. follow the CONSORT statement for reporting of randomised controlled trials (Shultz 2010);

2. follow a pre‐specified design as described in a protocol that is available in a publicly accessible registry;

3. use a study design with a sufficiently large sample size, adequate dosing, sufficient duration of intervention, and long‐term follow‐up, so that meaningful changes in cognitive scores can be assessed;

4. include outcomes that are important for people living with dementia;

5. include outcomes that will enable health economic analysis; and

6. include comparison with active comparators.

History

Protocol first published: Issue 10, 2017

Appendices

Appendix 1. Sources searched and search strategies

Source	Search strategy	Hits retrieved
1. CENTRAL (the Cochrane Library) http://crso.cochrane.org/SearchSimple.php (Date of most recent search: 8 July 2021)	#1 MESH DESCRIPTOR Dementia EXPLODE ALL TREES #2 MESH DESCRIPTOR Dementia EXPLODE ALL TREES #3 MESH DESCRIPTOR Delirium EXPLODE ALL TREES #4 MESH DESCRIPTOR Wernicke Encephalopathy EXPLODE ALL TREES #5 MESH DESCRIPTOR Neurocognitive Disorders EXPLODE ALL TREES #6 dement*:TI,AB,KY #7 alzheimer*:TI,AB,KY #8 (lewy* adj2 bod*):TI,AB,KY #9 (chronic adj2 cerebrovascular):TI,AB,KY #10 ("organic brain disease" or "organic brain syndrome"):TI,AB,KY #11 ("benign senescent forgetfulness"):TI,AB,KY #12 (cerebr* adj2 deteriorat*):TI,AB,KY #13 (cerebral* adj2 insufficient*):TI,AB,KY #14 #1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9 OR #10 OR #11 OR #12 OR #13 #15 MESH DESCRIPTOR Cannabinoids EXPLODE ALL TREES #16 MESH DESCRIPTOR Cannabis EXPLODE ALL TREES #17 MESH DESCRIPTOR Dronabinol EXPLODE ALL TREES #18 MESH DESCRIPTOR Endocannabinoids EXPLODE ALL TREES #19 MESH DESCRIPTOR Marijuana Smoking EXPLODE ALL TREES #20 MESH DESCRIPTOR Medical Marijuana EXPLODE ALL TREES #21 MESH DESCRIPTOR Receptors, Cannabinoid EXPLODE ALL TREES #22 9‐ene‐Tetrahydrocannabino:TI,AB,KY #23 Bhang:TI,AB,KY #24 cannabidiol:TI,AB,KY #25 Cannabinoid*:TI,AB,KY #26 Cannabis:TI,AB,KY #27 cbd:TI,AB,KY #28 Dronabinol:TI,AB,KY #29 endocannabinoid*:TI,AB,KY #30 Ganja:TI,AB,KY #31 Hashish:TI,AB,KY #32 Hemp:TI,AB,KY #33 Marihuana:TI,AB,KY #34 Marijuana:TI,AB,KY #35 Marinol:TI,AB,KY #36 Pot:TI,AB,KY #37 Tetrahydrocannabinol*:TI,AB,KY #38 THC:TI,AB,KY #39 #15 OR #16 OR #17 OR #18 OR #19 OR #20 OR #21 OR #22 OR #23 OR #24 OR #25 OR #26 OR #27 OR #28 OR #29 OR #30 OR #31 OR #32 OR #33 OR #34 OR #35 OR #36 OR #37 OR #38 #40 #14 AND #39	Sept 2017: 47 Jul 2018: 22 Jul 2019: 14 11 June 2020: 21 8 July 2021: 16
2. MEDLINE In‐process and other non‐indexed citations and MEDLINE 1950‐present (Ovid SP) (Date of most recent search: 8 July 2021)	1 exp Dementia/ 2 Delirium/ 3 Wernicke Encephalopathy/ 4 Delirium, Dementia, Amnestic, Cognitive Disorders/ 5 dement*.mp. 6 alzheimer*.mp. 7 (lewy* adj2 bod*).mp. 8 (chronic adj2 cerebrovascular).mp. 9 ("organic brain disease" or "organic brain syndrome").mp. 10 "benign senescent forgetfulness".mp. 11 (cerebr* adj2 deteriorat*).mp. 12 (cerebral* adj2 insufficient*).mp. 13 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or 12 14 exp Cannabinoids/ 15 exp Cannabis/ 16 exp Dronabinol/ 17 exp Endocannabinoids/ 18 exp Marijuana Smoking/ 19 exp Medical Marijuana/ 20 exp Receptors, Cannabinoid/ 21 9‐ene‐Tetrahydrocannabinol.ti,ab. 22 Bhang.ti,ab. 23 cannabidiol.ti,ab. 24 Cannabinoid*.ti,ab. 25 Cannabis.ti,ab. 26 "Cannabis indica".ti,ab. 27 "Cannabis sativa".ti,ab. 28 cbd.ti,ab. 29 "delta(1)‐Tetrahydrocannabinol".ti,ab. 30 "delta(1)‐THC".ti,ab. 31 "delta(9)‐Tetrahydrocannabinol".ti,ab. 32 "delta(9)‐THC".ti,ab. 33 "Delta‐9‐tetrahydrocannabinol".ti,ab. 34 Dronabinol.ti,ab. 35 endocannabinoid*.ti,ab. 36 Ganja.ti,ab. 37 Hashish.ti,ab. 38 Hemp.ti,ab. 39 Marihuana.ti,ab. 40 Marijuana.ti,ab. 41 Marinol.ti,ab. 42 Pot.ti,ab. 43 Tetrahydrocannabinol*.ti,ab. 44 THC.ti,ab. 45 or/14‐44 46 randomized controlled trial.pt. 47 controlled clinical trial.pt. 48 randomized.ab. 49 placebo.ab. 50 drug therapy.fs. 51 randomly.ab. 52 trial.ab. 53 groups.ab. 54 or/46‐53 55 13 and 45 and 54	Sept 2017: 230 Jul 2018: 40 Jul 2019: 49 11 June 2020: 50 8 July 2021: 64
3. Embase (Ovid SP 1974 to 2017 September 09) (Date of most recent search: 8 July 2021)	1 Dementia/ 2 Delirium/ 3 Wernicke Encephalopathy/ 4 Delirium, Dementia, Amnestic, Cognitive Disorders/ 5 ("benign senescent forgetfulness" or ("normal pressure hydrocephalus" and "shunt*") or ("organic brain disease" or "organic brain syndrome") or ((cerebral* or cerebrovascular or cerebro‐vascular) adj2 insufficien*) or (cerebr* adj2 deteriorat*) or (chronic adj2 (cerebrovascular or cerebro‐vascular)) or (creutzfeldt or jcd or cjd) or (lewy* adj2 bod*) or (pick* adj2 disease) or alzheimer* or binswanger* or deliri* or dement* or huntington* or korsako*).tw. 6 or/1‐5 7 exp Cannabinoids/ 8 exp Cannabis/ 9 exp Dronabinol/ 10 exp Endocannabinoids/ 11 exp Marijuana Smoking/ 12 exp Medical Marijuana/ 13 exp Receptors, Cannabinoid/ 14 9‐ene‐Tetrahydrocannabinol.ti,ab. 15 Bhang.ti,ab. 16 cannabidiol.ti,ab. 17 Cannabinoid*.ti,ab. 18 Cannabis.ti,ab. 19 "Cannabis indica".ti,ab. 20 "Cannabis sativa".ti,ab. 21 cbd.ti,ab. 22 "delta(1)‐Tetrahydrocannabinol".ti,ab. 23 "delta(1)‐THC".ti,ab. 24 "delta(9)‐Tetrahydrocannabinol".ti,ab. 25 "delta(9)‐THC".ti,ab. 26 "Delta‐9‐tetrahydrocannabinol".ti,ab. 27 Dronabinol.ti,ab. 28 endocannabinoid*.ti,ab. 29 Ganja.ti,ab. 30 Hashish.ti,ab. 31 Hemp.ti,ab. 32 Marihuana.ti,ab. 33 Marijuana.ti,ab. 34 Marinol.ti,ab. 35 Pot.ti,ab. 36 Tetrahydrocannabinol*.ti,ab. 37 THC.ti,ab. 38 or/7‐37 39 randomized controlled trial/ 40 controlled clinical trial/ 41 random$.ti,ab. 42 randomization/ 43 intermethod comparison/ 44 placebo.ti,ab. 45 (compare or compared or comparison).ti. 46 ((evaluated or evaluate or evaluating or assessed or assess) and (compare or compared or comparing or comparison)).ab. 47 (open adj label).ti,ab. 48 ((double or single or doubly or singly) adj (blind or blinded or blindly)).ti,ab. 49 double blind procedure/ 50 parallel group$1.ti,ab. 51 (crossover or cross over).ti,ab. 52 ((assign$ or match or matched or allocation) adj5 (alternate or group$1 or intervention$1 or patient$1 or subject$1 or participant$1)).ti,ab. 53 (assigned or allocated).ti,ab. 54 (controlled adj7 (study or design or trial)).ti,ab. 55 (volunteer or volunteers).ti,ab. 56 trial.ti. 57 or/39‐56 58 6 and 38 and 57	Sept 2017: 427 Jul 2018: 135 Jul 2019: 110 11 June 2020: 114 8 July 2021: 104
4. PsycINFO (Ovid SP) (Date of most recent search: 8 July 2021)	1 exp Dementia/ 2 exp Delirium/ 3 exp Huntingtons Disease/ 4 exp Kluver Bucy Syndrome/ 5 exp Wernickes Syndrome/ 6 exp Cognitive Impairment/ 7 dement*.mp. 8 alzheimer*.mp. 9 (lewy* adj2 bod*).mp. 10 deliri*.mp. 11 (chronic adj2 cerebrovascular).mp. 12 ("organic brain disease" or "organic brain syndrome").mp. 13 "supranuclear palsy".mp. 14 ("normal pressure hydrocephalus" and "shunt*").mp. 15 "benign senescent forgetfulness".mp. 16 (cerebr* adj2 deteriorat*).mp. 17 (cerebral* adj2 insufficient*).mp. 18 (pick* adj2 disease).mp. 19 (creutzfeldt or jcd or cjd).mp. 20 huntington*.mp. 21 binswanger*.mp. 22 korsako*.mp. 23 ("parkinson* disease dementia" or PDD or "parkinson* dementia").mp. 24 or/1‐23 25 exp CANNABINOIDS/ 26 exp CANNABIS/ 27 exp Tetrahydrocannabinol/ 28 exp Marijuana/ 29 exp Marijuana Usage/ 30 "9‐ene‐Tetrahydrocannabinol".ti,ab. 31 Bhang.ti,ab. 32 cannabidiol.ti,ab. 33 Cannabinoid*.ti,ab. 34 Cannabis.ti,ab. 35 "Cannabis indica".ti,ab. 36 "Cannabis sativa".ti,ab. 37 cbd.ti,ab. 38 "delta(1)‐Tetrahydrocannabinol".ti,ab. 39 "delta(1)‐THC".ti,ab. 40 "delta(9)‐Tetrahydrocannabinol".ti,ab. 41 "delta(9)‐THC".ti,ab. 42 "Delta‐9‐tetrahydrocannabinol".ti,ab. 43 Dronabinol.ti,ab. 44 endocannabinoid*.ti,ab. 45 Ganja.ti,ab. 46 Hashish.ti,ab. 47 Hemp.ti,ab. 48 Marihuana.ti,ab. 49 Marijuana.ti,ab. 50 Marinol.ti,ab. 51 Pot.ti,ab. 52 Tetrahydrocannabinol*.ti,ab. 53 THC.ti,ab. 54 or/25‐53 55 exp Clinical Trials/ 56 randomly.ab. 57 randomi?ed.ti,ab. 58 placebo.ti,ab. 59 groups.ab. 60 "double‐blind*".ti,ab. 61 "single‐blind*".ti,ab. 62 RCT.ti,ab. 63 or/55‐62 64 24 and 54 and 63	Sept 2017: 103 Jul 2018: 12 Jul 2019: 9 11 June 2020: 8 8 July 2021: 11
5. CINAHL (EBSCOhost) (Date of most recent search: 8 July 2021)	1 MH "Dementia+" 2 MH "Delirium") or (MH "Delirium, Dementia, Amnestic, Cognitive Disorders") 3 MH "Wernicke's Encephalopathy" 4 TX dement* 5 TX alzheimer* 6 TX lewy* N2 bod* 7 TX deliri* 8 TX chronic N2 cerebrovascular 9 TX "organic brain disease" or "organic brain syndrome" 10 TX "normal pressure hydrocephalus" and "shunt*" 11 TX "benign senescent forgetfulness" 12 TX cerebr* N2 deteriorat* 13 TX cerebral* N2 insufficient* 14 TX pick* N2 disease 15 TX creutzfeldt or jcd or cjd 16 TX huntington* 17 TX binswanger* 18 TX korsako* 19 (S1 OR S2 OR S3 OR S4 OR S5 OR S6 OR S7 OR S8 OR S9 OR S10 OR S11 OR S12 OR S13 OR S14 OR S15 OR S16 OR S17 OR S18) 20 (MH "Cannabis") 21 (MH "Medical Marijuana") 22 TX Bhang 23 TX cannabidiol 24 TX Cannabinoid* 25 TX Cannabis 26 TX cbd 27 TX "delta(9)‐Tetrahydrocannabinol" 28 TX "delta(9)‐THC" 29 TX "Delta‐9‐tetrahydrocannabinol" 30 TX Dronabinol 31 TX endocannabinoid* 32 TX Ganja 33 TX Hashish 34 TX Hemp 35 TX Marihuana 36 TX Marijuana 37 TX Marinol 38 TX Pot 39 TX Tetrahydrocannabinol* 40 TX THC 41 S20 OR S21 OR S22 OR S23 OR S24 OR S25 OR S26 OR S27 OR S28 OR S29 OR S30 OR S31 OR S32 OR S33 OR S34 OR S35 OR S36 OR S37 OR S38 OR S39 OR S40 42 MH "Clinical Trials" 43 TX trial 44 TX "single‐blind*" 45 TX "double‐blind*" 46 TX "treatment as usual" 47 TX randomly 48 S42 OR S43 OR S44 OR S45 OR S46 OR S47 49 S19 AND S41 AND S48	Sept 2017: 40 Jul 2018: 11 Jul 2019: 14 11 June 2020: 8 8 July 2021: 20
6. ISI Web of Science – all databases [includes Web of Science (1945‐present); BIOSIS Previews (1926‐present); MEDLINE (1950‐present); Journal Citation Reports] (Date of most recent search: 8 July 2021)	(Cannabinoids OR Cannabis OR Dronabinol OR Endocannabinoids OR Marijuana OR Marihuana OR Tetrahydrocannabinol) AND TOPIC:(dement* OR alzheimer* OR "vascular cognitive impairment" OR "lew* bod*" OR CADASIL OR "cognit* impair*" OR FTD OF FTLD OR "cerebrovascular insufficienc*" OR AD OR VCI) ANDTOPIC: (randomly OR randomised OR randomized OR "random allocat*" OR RCT OR CCT OR "double blind*" OR "single blind*" OR "double blind*" OR "single blind*" OR trial)	Sept 2017: 210 Jul 2018: 27 Jul 2019: 31 11 June 2020: 45 8 July 2021: 299
7. LILACS (BIREME) (Date of most recent search: 8 July 2021)	Cannabinoids OR Cannabis OR Dronabinol OR Endocannabinoids OR Marijuana OR Marihuana OR Tetrahydrocannabinol [Words] and alzheimer OR alzheimers OR alzheimer’s OR dementia OR demenc$ [Words] and randomly OR randomised OR randomized OR RCT OR "controlled trial" OR "double blind$" OR placebo	Sept 2017: 0 Jul 2018: 0 Jul 2019: 0 11 June 2020: 0 8 July 2021: 1
8. ClinicalTrials.gov (www.clinicaltrials.gov) (Date of most recent search: 8 July 2021)	dementia OR alzheimers OR cognition OR cognitive | Cannabinoids OR Cannabis OR Dronabinol OR Endocannabinoids OR Marijuana OR Marihuana OR Tetrahydrocannabinol	Sept 2017: 40 Jul 2018: 13 Jul 2019: 8 11 June 2020: 8 8 July 2021: 14
9. ICTRP (Date of most recent search: 8 July 2021)	dementia OR alzheimers OR cognition OR cognitive | Cannabinoids OR Cannabis OR Dronabinol OR Endocannabinoids OR Marijuana OR Marihuana OR Tetrahydrocannabinol	Sept 2017: 280 Jul 2018: 9 Jul 2019: 14 11 June 2020: 0 8 July 2021: 8
10. ALOIS (CDCIG Specialised Register: www.medicine.ox.ac.uk/alois) (Date of most recent search: 8 July 2021)	Cannabinoids OR Cannabis OR Dronabinol OR Endocannabinoids OR Marijuana OR Marihuana OR Tetrahydrocannabinol	Sept 2017:2 Jul 2018: 0 Jul 2019: 4 11 June 2020: 6 8 July 2021: 19
TOTAL before de‐duplication		TOTAL: 2,936
TOTAL after de‐duplication		TOTAL: 2,448

Data and analyses

Comparison 1. Comparison: cannabinoids versus placebo.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Changes in global and specific cognitive function, sMMSE scale	1		Mean Difference (IV, Fixed, 95% CI)	1.10 [0.10, 2.10]
1.2 Overall behavioural and psychological symptoms of dementia, NPI  and NPI‐NH total score	3		Mean Difference (IV, Fixed, 95% CI)	‐1.97 [‐3.87, ‐0.07]
1.3 Subgroup analysis ‐ Overall behavioural and psychological symptoms of dementia, NPI  and NPI‐NH total score	3		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
1.3.1 Other studies ‐ NPI, less severely ill, oral THC	2		Mean Difference (IV, Fixed, 95% CI)	0.01 [‐2.51, 2.52]
1.3.2 Herrmann ‐ NPI‐NH, more severely ill, synthetic THC analogue	1		Mean Difference (IV, Fixed, 95% CI)	‐4.60 [‐7.50, ‐1.70]
1.4 Adverse events ‐ nervous system disorders	1		Odds Ratio (M‐H, Fixed, 95% CI)	Subtotals only
1.4.1 Total nervous system disorders	1	50	Odds Ratio (M‐H, Fixed, 95% CI)	0.71 [0.23, 2.18]
1.4.2 Dizziness	1	50	Odds Ratio (M‐H, Fixed, 95% CI)	1.10 [0.24, 4.99]
1.4.3 Somnolence	1	50	Odds Ratio (M‐H, Fixed, 95% CI)	0.50 [0.08, 3.02]
1.4.4 Muscle spasm	1	50	Odds Ratio (M‐H, Fixed, 95% CI)	0.35 [0.01, 8.93]
1.5 Adverse events ‐ Sedation/lethargy	1		Odds Ratio (IV, Fixed, 95% CI)	2.83 [1.07, 7.48]
1.6 Adverse events ‐ treatment‐induced sedation	1		Odds Ratio (IV, Fixed, 95% CI)	4.01 [0.40, 40.56]
1.7 Adverse events ‐ psychiatric disorders	1		Odds Ratio (M‐H, Fixed, 95% CI)	Subtotals only
1.7.1 Total psychiatric disorders	1	50	Odds Ratio (M‐H, Fixed, 95% CI)	2.26 [0.57, 9.02]
1.7.2 Euphoria	1	50	Odds Ratio (M‐H, Fixed, 95% CI)	0.35 [0.01, 8.93]
1.8 Adverse events ‐ gastrointestinal disorders	1		Odds Ratio (M‐H, Fixed, 95% CI)	Subtotals only
1.8.1 Total gastrointestinal disorders	1	50	Odds Ratio (M‐H, Fixed, 95% CI)	2.40 [0.40, 14.49]
1.8.2 Nausea	1	50	Odds Ratio (M‐H, Fixed, 95% CI)	2.27 [0.19, 26.81]
1.9 Adverse events ‐ other	1		Odds Ratio (M‐H, Fixed, 95% CI)	Subtotals only
1.9.1 Total adverse events‐ other	1	50	Odds Ratio (M‐H, Fixed, 95% CI)	0.70 [0.11, 4.58]
1.9.2 Fatigue	1	50	Odds Ratio (M‐H, Fixed, 95% CI)	1.09 [0.14, 8.42]
1.10 Changes in functional outcomes, Barthel Index	1		Mean Difference (IV, Fixed, 95% CI)	0.60 [‐0.75, 1.95]
1.11 Minimal to marked improvement on dementia severity, assessed on the CGI Change scale	2		Odds Ratio (IV, Fixed, 95% CI)	1.88 [1.03, 3.44]
1.12 Weight [kg]	3		Mean Difference (IV, Fixed, 95% CI)	0.33 [‐0.08, 0.75]
1.13 MNA‐SF	1		Mean Difference (IV, Fixed, 95% CI)	0.20 [0.02, 0.38]
1.14 BMI	1		Mean Difference (IV, Fixed, 95% CI)	‐0.14 [‐0.35, 0.07]
1.15 Caloric intake [kcal]	1		Mean Difference (IV, Fixed, 95% CI)	19.00 [‐508.74, 546.74]
1.16 CMAI	3		Mean Difference (IV, Fixed, 95% CI)	‐2.35 [‐4.10, ‐0.60]
1.17 NPI subscale agitation/aggression	3		Mean Difference (IV, Fixed, 95% CI)	‐0.63 [‐1.08, ‐0.18]
1.18 QoL‐AD	1		Mean Difference (IV, Fixed, 95% CI)	‐0.50 [‐2.60, 1.60]
1.19 Caregiver burden	2		Std. Mean Difference (IV, Fixed, 95% CI)	‐0.12 [‐0.38, 0.13]
1.20 All‐cause discontinuation	2	126	Odds Ratio (M‐H, Fixed, 95% CI)	1.02 [0.33, 3.13]
1.21 All‐cause mortality	2	100	Odds Ratio (M‐H, Fixed, 95% CI)	0.59 [0.07, 4.62]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Herrmann 2019.

Study characteristics
Methods	Type of RCT: randomised, double‐blind, placebo‐controlled, cross‐over trial Method of allocation: cross‐over assignment, masking‐quadruple (participant, care provider, investigator, outcomes assessor) Date: study start date: January, 2015; actual primary completion date: January 2018 Study duration: 3 years (duration of intervention: 14 weeks ‐ 6 weeks of each intervention preceded by 1 week of washout)
Participants	Location: Sunnybrook Health Sciences Centre, Canada Geographical area: Toronto, Ontario, Canada Inclusion criteria: males or females ≥ 55 years of age; Diagnostic and Statistical Manual (DSM) V criteria for Major Neurocognitive Disorder due to AD or both Major Neurocognitive Disorder due to AD and Major Vascular Neurocognitive Disorder (i.e. mixed AD and cerebrovascular disease); currently in moderate to severe stage of dementia (Mini‐Mental Status Examination ≤ 24); presence of clinically significant agitation (Neuropsychiatric Inventory (NPI) agitation subscale ≥ 3); in case of treatment with cognitive‐enhancing medications (cholinesterase inhibitors and/or memantine), dosage must be stable for at least 3 months, or if the ChEI and/or memantine has been discontinued, patients may enrol after 1 month Exclusion criteria: change in psychotropic medications less than 1 month before study randomisation (e.g. concomitant antidepressants); contraindications to nabilone (history of hypersensitivity to any cannabinoid); current or past significant cardiovascular disease (e.g. uncontrolled hypertension, ischaemic heart disease, arrhythmia and severe heart failure); presence or history of other psychiatric disorders or neurological conditions (e.g. psychotic disorders, schizophrenia, stroke, epilepsy); previous or current abuse of/dependence on marijuana Sample: 39 randomised participants Age: ≥ 55 years Sex: all Baseline characteristics: out of 39 patients, 28 were institutionalised and 11 were in outpatient care  Baseline values of the specific tests (mean ± SD): sMMSE: 6.5 ± 6.8  NPI‐NH Total: 34.3 ± 15.8  CGIC: 3.7 ± 0.9  CMAI: 67.9 ± 17.6 MNA‐SF: 8.7 ± 2.9
Interventions	Intervention: nabilone (∆9‐tetrahydrocannabinol (THC) analogue) Comparator: placebo Intervention delivery: after a 1‐week placebo washout, participants randomised to nabilone treatment arm were administered nabilone treatment (Weeks 1 to 6) in daily doses od 0.5 to maximum target dose of 2 mg. In Week 1, participants received 0.25 mg before bedtime for 3 nights, then were titrated up to 0.5 mg/d total for the next 4 days. In Week 2, doses were titrated up to 1 mg/d, and during Weeks 3 and 4, doses were flexible (1 to 2 mg/d based on tolerability). The tolerated dose was maintained until down‐titration in Week 6. During the last 4 days of Week 6, doses were tapered down to reduce the risk of potential withdrawal. On Days 6 and 7 of Week 6, participants received 0.5 mg/d total. After 6 weeks, they received a 1‐week placebo washout (Week 7) before receiving 6 weeks of placebo treatment (Weeks 8 to 14). Participants randomised to the placebo arm had a 1‐week placebo washout, followed by 6 weeks of drug‐matched placebo (Weeks 1 to 6). The placebo phase followed the same dosing structure as was described for the nabilone arm. During Week 7, they received a 1‐week placebo washout (Week 7) before receiving 6 weeks of nabilone treatment (Weeks 8 to 14). Setting: a long‐term care facility and geriatric psychiatry clinics
Outcomes	In both protocol and abstracts: Primary Outcome Measures ‐ Change in agitation: Cohen‐Mansfield Agitation Inventory (CMAI) Secondary Outcome Measures ‐ Change in neuropsychiatric symptoms: Neuropsychiatric Inventory ‐ Nursing Home (NPI‐NH) ‐ Change in cognition: Standardized Mini‐Mental State Examination (sMMSE), Severe Impairment Battery (SIB), Alzheimer's Disease Assessment Scale ‐ Cognitive (ADAS‐Cog) ‐ Change in clinical representation: Alzheimer's Disease Cooperative Study ‐ Clinician Global Impression (ADCS ‐ CGIC) Other Outcome Measures ‐ Change in pain: Pain Assessment In Advanced Dementia (PAINAD) ‐ Change in nutritional status: Mini Nutritional Assessment ‐ Short Form (MNA‐SF) ‐ Change in heart rate, levels of blood biomarkers, blood pressure
Notes	Funding: sponsors and collaborators ‐ Sunnybrook Health Sciences Centre Trial registration: Number of protocol on ClinicalTrials.gov: NCT02351882 Other study ID numbers: 318‐2013
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "a block randomization code was independently computer generated by the institutional Pharmacy Department."
Allocation concealment (selection bias)	Low risk	Study authors reported that allocation was concealed, but they did not report the exact method of allocation concealment in the text of the published manuscript. Quote: "patients, family members, nurses, clinicians, outcome assessors, and investigators were blinded to treatment allocation and block size" Based on communication with the principal investigator, the following additional information was obtained, indicating that an adequate method of allocation concealment was used (sequentially numbered, opaque, sealed envelopes). Quote: "the randomization code for each participant by randomization number was kept in a sequentially numbered sealed opaque envelope under double‐key in case emergency unblinding was required. No emergency unblinding was done. Allocation code was only released from the Pharmacy once all other data entry was completed and verified, and the database was locked."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blind, with unblinding in exceptional occasions. Quote: "study staff were unblinded only when the final patient completed all study assessments and the database was locked. Unblinding during the study was permitted only in exceptional clinical circumstances (i.e. necessary for acute medical management of serious adverse events [SAEs]) and only after approval by the principal or qualified investigator (PI or QI)." It was transparently reported which individuals were blinded. Quote: "patients, family members, nurses, clinicians, outcome assessors, and investigators were blinded to treatment allocation and block size."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Assessment was performed by trained study staff. Nurses, clinicians, outcome assessors, and investigators were blinded to treatment allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Out of 39 enrolled participants, one discontinued the study during the first week (placebo) due to clinically significant delusions. Therefore, 38 participants were included in the statistical analyses. Out of 38 participants, two died before study completion, and nine were terminated early due to serious adverse events (five during the nabilone phase and four during placebo phase).
Selective reporting (reporting bias)	Unclear risk	The study protocol was registered at www.clinicaltrials.gov (clinical trial identifier number NCT02351882). All outcomes mentioned in the study protocol were reported.
Other bias	Low risk	Other sources of bias were not found. We consider the washout period to have been adequate for avoidance of carry‐over effects.

van den Elsen NCT01302340.

Study characteristics
Methods	Type of RCT: Multicenter, crossover; study of effectiveness; superiority trial Method of allocation: Multicenter, crossover; study of effectiveness; superiority trial Date: September 2011 to December 2013 (Note: Date was not specified in the manuscript Ahmed 2015) Study duration: 12 weeks
Participants	Location: Radboud UMC Alzheimer Centre (Nijmegen, The Netherlands) and the Vincent van Gogh Institute for Psychiatry (Venray, The Netherlands) Geographical area: Nijmegen and Venray (The Netherlands) (Note: Location specified in manuscript Ahmed 2015 was Radboud University Medical Center, Nijmegen, Netherlands) Inclusion criteria: · Diagnosis of Alzheimer's Disease (AD), Vascular Dementia (VD) or mixed, according to the criteria of NINCDS‐ADRDA or NINCDS‐AIREN · Clinical Dementia Rating score between 0.5 and 3 · NPS symptoms, with at least agitation or aggression (Neuropsychiatric Inventory [NPI] score ≥10) · Informal caregiver available (Note: Additional inclusion criteria for van den Elsen, 2017: · being able to walk at least 10 m and understand simple instructions) Exclusion criteria: · Diagnosis of Lewy Body Dementia (LBD) or Fronto‐Temporal Dementia (FTD) · Major psychiatric disorder · Severe concomitant illness, seizure, arrhythmias (except sinus arrhythmia and atrial fibrillation), heart failure New York Heart Association (NYHA) class III or IV · Frequent falling due to orthostatic hypotension · History of alcohol or drug abuse · Tri Cyclic Antidepressives (TCA) or opioids used within 30 days before randomization till the end of the study · Changes in dosage of antidepressives within 6 weeks before randomization and during study, and changes in dosage antipsychotics or benzodiazepines within 2 weeks prior to randomization and during study · Drugs from a predesigned list of inhibitors of cytochrome P‐450 enzymes Sample: 22 randomized patients Age: mean age 76.4 (5.3) years Sex: 15 men, 7 women (Note: In van den Elsen 2015, the total number of participants was reported as 18, mean age 77, 15 men, 3 women. In the manuscript Ahmed, 2015, the total number of participants analysed was 10, mean age 77.3±5.6 years, 7 men and 3 women.) Baseline characteristics: Patients with dementia and clinically relevant NPS. Out of 22 included participants, 12 were from the "ambulatory group" and 10 from the "hospital admission group", with significantly higher baseline NPI scores in the "ambulatory group" compared to the hospital admission group (t= ‐2.56, df=20, p=0.019). Baseline values of the specific tests (mean ± SD): sMMSE: Total: 16.9 ± 7.8; Hospital group: 18.5 ± 6.0; Ambulatory group: 15.2 ± 9.2 NPI: Total: 35.0 ± 16.5; Hospital group: 26.2 ± 17.7; Ambulatory group: 42.3 ± 11.8 CMAI: Total: 58.3 ± 17.4; Hospital group: 54.0 ± 19.8; Ambulatory group: 62.0 ± 15.1 ZBI: Total: 36.0 ± 15.1; Hospital group: 34.5 ± 19.0; Ambulatory group: 37.3 ± 11.4
Interventions	Intervention: tetrahydrocannabinol (THC) Comparator: placebo Intervention delivery: Within each block THC (0.75mg twice daily in block 1‐3 and 1.5mg twice daily in block 4‐6) and placebo were administered in random order for three consecutive days, followed by four‐day washout. Each group consisted of 11 patients. Frequency: Twice daily for 3 days, separated by a 4‐day washout period during 12 weeks. Setting: Two hospital sites: Radboud university medical center (Nijmegen, Gelderland, Netherlands) and Vincent van Gogh Institute for Psychiatry, department of Elderly (Venray, Limburg, Netherlands), from September 2011 to December 2013. Supervision: Primary caregiver, investigator. Adherence: High; 98.5% of tablets were administered (THC, 99.5%; placebo, 97.8%). (Note: The manuscript Ahmet, 2015 described adherence of almost 98 %, THC 99 %; placebo 97.5 %)
Outcomes	Manuscript Ahmed 2015: Safety and tolerability (adverse effects, incidence and severity): · reported by patients and caregivers, observed clinically (time of onset, severity, duration, and causal relationship to study drugs) · physical examination · vital signs · 12‐lead electrocardiogram · laboratory tests (haematology and clinical chemistry) · causality assessed by a research physician, five‐point scale: (1) unrelated, (2) unlikely, (3) possibly (4), probably, (5) or definitely related to the intervention · existence of serious adverse event (fatal or life‐threatening, that required (prolonged) hospitalizations, or that resulted in persistent or significant disability or incapacity) Pharmacodynamic effects: · psychedelic effects: the Bowdle VAS for psychedelic (Bowdle et al. 1998; Zuurman et al. 2008) · body sway: with the SwayStarTM, a wireless device attached to the trunk (http://www.b2i.info/web/index.htm) · vital signs: systolic and diastolic blood pressure and heart rate Pharmacokinetic parameters: · quantification of THC and 11‐OH‐THC (liquid chromatography with tandem‐mass spectrometer detection) · terminal half‐life (t1/2), area under the curve (AUC) from 0 to 24 h (AUC0–24 h), and apparent clearance (CL/F, being the dose/AUC0–24 h), Phoenix WinNonlin software version 6.3 (Certara, L.P./Pharsight Ltd) Manuscript van den Elsen 2015: Primary outcome: · Neuropsychiatric Inventory (NPI) Secondary outcomes: · Cohen‐Mansfield Agitation Inventory (CMAI) · Zarit Burden Scale (ZBS) · Visual Analogue Scale (VAS) Bowdle for feeling high · Gait rite® · Sway Star® · Time Up and Go (TUG) · Tinetti Performance–Oriented Mobility Assessment (Tinetti POMA) Other assessments of safety: vital signs, physical examination and weight, laboratory tests, electrocardiography and Delirium Observation Scale
Notes	Funding: Funding was provided by the European Regional Development Fund (ERDF) and the Province of Gelderland, which both had no role in study design, collection, analysis, interpretation of data, nor writing the report. Declaration of interest: GE, TF and MM were supported by a grant of the ERDF for the conduct of the current study. The other authors have no disclosures to report. Trial registration: ClinicalTrials.gov, NCT01302340
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "The randomization codes were generated by an independent pharmacist, using a computer algorithm for random numbers."
Allocation concealment (selection bias)	Low risk	In the published article, the authors reported that allocation was concealed, but they did not report exact method of allocation concealment. The authors provided the following information regarding allocation concealment method via e‐mail:  "Subjects who signed informed consent received a screening number starting with the letter “S”. Treatment allocation was done on basis of eligibility of the subject and subject numbers were assigned at that moment. In other words, after screening each eligible subject received a subject number in consecutive order, starting with the lowest number available. The randomization was double‐blind, with dose incrementation. Randomization was carried out by the pharmacy of UMC St Radboud using MS Excel. The pharmacist delivered the study medication on the subject’s number, without reference to the content. Allocation was only present at this pharmacist, who was only contacted again after closing of the trial master file en the data acquisition and registration in Castor and after the analyses plan was agreed on." Based on this response, we concluded that the central allocation concealment method was used, via a third person who was not involved in other segments of the trial.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	In addition to describing the study as "double‐blind", the authors also reported: "Treatment allocation was strictly concealed from participants, caregivers, investigators, and all other personnel directly involved in the study and was not made available until study completion and database lock." This was interpreted as referring to post‐allocation concealment (i.e. blinding).
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "The causality was assessed by a research physician, blinded to treatment allocation."
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "All participants completed the study as scheduled."
Selective reporting (reporting bias)	Low risk	The study protocol was registered on www.clinicaltrials.gov (clinical trial identifier number NCT01302340). All outcomes mentioned in the study protocol were reported.
Other bias	Low risk	Other sources of bias not found. We consider the washout period to have been adequate for the avoidance of carry over effects.

van den Elsen NCT01608217.

Study characteristics
Methods	Type of RCT: randomised, double‐blind, placebo‐controlled study; multi‐centre, phase II trial Method of allocation: parallel assignment, masking‐quadruple (participant, care provider, investigator, outcomes assessor) Date: study start date: June 2012; study completion date: June 2014 Study duration: 5 weeks Duration of therapy: 3 weeks Follow‐up assessments by telephone were performed 2 weeks after study completion
Participants	Location: Netherlands Participants were recruited from 9 participating institutes throughout the southeast of The Netherlands. Geriatric outpatient clinics (n = 2 clinics). Psychiatric clinics (n = 3). Nursing homes (n = 3, including in total 6 locations). Regional network of integrated care for community‐dwelling patients with dementia. Geographical area: The Netherlands Inclusion criteria Subject has possible or probable dementia, type Alzheimer, vascular or mixed‐type dementia, according to the criteria of NINCDS‐ADRDA/NINCDS‐AIREN, or based on an expert panel decision Clinical Dementia Rating (CDR) score 1 to 3 (mild to severe dementia) Age ≥ 40 years Clinically relevant behavioural disturbances occurring at least 1 month before screening, defined as score ≥ 10 on the NPI, including presence of the domain agitation/aggression or motor disturbance Women should be in the postmenopausal phase Availability of an informal or formal caregiver, who is in touch with the patient at least twice a week Informed consent by the patient and the patient's informal caregiver If applicable: patient is willing to stop his/her own pain medication for the duration of the study Exclusion criteria Dementia other than AD, VaD, or AD/VaD Major psychiatric disorder such as: major depression according to DSM IV within 6 months before randomisation, history of psychosis or mania, current hallucinations and/or delirium, current suicidal ideation, or major anxiety disorder History of, or current, drug abuse Current alcohol abuse or unwillingness to use no more than 2 alcoholic consumptions daily or raised gamma‐glutamine transpeptidase and alkaline phosphatase Clinical or biochemical evidence of liver disease (ALT or AST ≥ twice the upper limit of normal) or known allergy to acetaminophen Severe (and/or unstable) concomitant or intercurrent illness, such as seizure, arrhythmias requiring drugs other than a beta‐blocker or digoxin (except sinus arrhythmia and atrial fibrillation), unstable angina pectoris, heart failure NYHA III or IV, severe concomitant illness that requires treatment changes Known or suspected sensitivity to cannabinoids Lactose intolerance Frequent falling due to orthostatic hypotension Use of tricyclic antidepressants (TCAs), fluoxetine, and/or carbamazepine Change in dosage of antipsychotics, benzodiazepines, or cholinesterase inhibitors within 2 weeks before intervention Participation in any study other than the descriptive 'Parelsnoer' study Sample: 54 patients were assessed for eligibility, 50 of whom were randomised and received study medication (THC, n = 24; placebo, n = 26) Age: 40 years and older (adult, senior) Sex: 25 men, 25 women Baseline characteristics: patients with diagnosed AD or vascular or mixed dementia according to National Institute of Neurological and Communicative Disorders and Stroke–Alzheimer’s Disease and Related Disorders Association or National Institute of Neurological Disorders and Stroke–Association Internationale pour la Recherche en l’Enseignement en Neurosciences criteria were eligible for participation if they had clinically relevant NPS (minimal Neuropsychiatric Inventory [NPI] score ≥ 10), with symptoms of agitation, aggression, or aberrant motor behaviour existing at least 1 month before screening  Patients were recruited from community dwelling (total: n = 24, 48.0%; delta‐THC: n = 13, 54.2%; placebo: n = 11, 42.3%), specialised dementia care unit (total: n = 13, 26.0%; delta‐THC: n = 4, 16.7%; placebo: n = 9, 34.6%), and nursing homes (total: n = 13, 26.0%; delta‐THC: n = 7, 29.2%; placebo: n = 6, 23.1%) Baseline values of the specific tests (mean ± SD) sMMSE: total: 14.8 ± 6.7; delta‐THC: 15.9 ± 6.7, placebo: 14.0 ± 6.8 NPI: delta‐THC: 37.4 ± 13.7; placebo: 35.6 ± 13.0 CMAI: delta‐THC: 58.8 ± 18.5; placebo: 61.6 ± 16.4 Barthel Index: delta‐THC: 13.8 ± 5.1; placebo: 13.3 ± 5.3 QoL‐AD: delta‐THC: 28.3 ± 4.9; placebo: 29.6 ± 5.2 CCGIC: delta‐THC: 3.7 ± 1.0; placebo: 3.4 ± 1.2
Interventions	Intervention: delta‐9‐tetrahydrocannabinol (delta‐THC) (N = 24) Comparator: placebo (N = 26) Intervention schedule: delta‐9‐tetrahydrocannabinol (delta‐THC) 1.5 mg (tablet) 3 times daily for a period of 3 weeks. Other Names of the intervention: Namisol (Echo Pharmaceuticals, Weesp, The Netherlands), Cannabis, ECP002A. placebo (tablet) 3 times daily for a period of 3 weeks Additionally, patients received 1000 mg acetaminophen 3 times daily in case of pain complaints, or of suspected pain in non‐communicative patients, based on physical examination at screening and information from the caregiver or physician Setting: assessments were done by researchers from the Department of Geriatric Medicine of Radboudumc (Nijmegen, The Netherlands) and the Department of Elderly of Vincent van Gogh Institute (psychiatric hospital, Venray, The Netherlands) Supervision: efficacy assessments were scheduled after 14 ± 2 treatment days (phone call) and 21 ± 2 treatment days (visit). For the purposes of safety assessment and compliance, several phone calls were performed by researchers during the intervention period (Days 2, 7, and 14). Follow‐up assessments by telephone were performed 2 weeks after study completion Adherence: median treatment compliance, based on remaining pill count, was 98% (67% to 100%) in the THC group and 100% (94% to 100%) in the placebo group
Outcomes	Primary outcome measure Change in NPS, measured with NPI Secondary efficacy outcome measures In protocol and manuscript: Secondary outcomes Included assessment of agitated behaviour and aggression (Cohen‐ Mansfield Agitation Inventory (CMAI)), activities of daily living (Barthel Index), and quality of life (Quality of Life ‐ Alzheimer’s Disease Scale (QoL‐AD)). All were assessed at baseline and at Day 21. Overall change was assessed by the primary caregiver, using the Caregiver Clinical Global Impression of Change (CCGIC), a 7‐point scale ranging from marked improvement to marked worsening from baseline. The Paired Associate Learning Wechsler Memory Scale‐Revised (PAL WMS‐R) was used for assessment of possible effects of THC on episodic memory function (baseline and Day 21) Secondary outcomes described in Appendix e‐1 and Table e‐1 on the Neurology® website at Neurology.org Pain Assessment Checklist for Seniors with Limited Ability to Communicate Dutch version (PACSLAC‐D) [Time Frame: baseline (T = 0) and T = 3 weeks] Verbal Rating Scale (VRS) [Time Frame: screening, baseline, T = 3 weeks, follow‐up (T = 5 weeks) and daily in a medication diary]
Notes	Funding: "study was funded by the European Regional Development Fund and the Province of Gelderland. The Investigation Medical Product was provided by Echo Pharmaceuticals, Weesp, The Netherlands, which did not provide financial support for the study and had no role in study design, collection, analysis or interpretation of data, or writing the report" Declaration of interest: G. van den Elsen was supported by a grant from the European Regional Development Fund for the conduct of the study. A. Ahmed, R. Verkes, and C. Kramers report no disclosures relevant to the manuscript. T. Feuth was supported by a grant from the European Regional Development Fund for the conduct of the study. P. Rosenberg reports no disclosures relevant to the manuscript. M. van der Marck was supported by a grant from the European Regional Development Fund for the conduct of the study. M. Olde Rikkert reports no disclosures relevant to the manuscript Trial registration Number of protocol at ClinicalTrials.gov: NCT01608217 Other Study ID Numbers: GER001‐02‐02 2011‐005289‐39 (EudraCT Number) NL38617.091.12 (Registry Identifier: toetsingonline.com)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "randomization (allocation ratio 1:1) was performed by an independent statistician using a computer‐generated randomization program, of which the algorithm was stratified per center and minimized for NPI score, dementia severity, sex, and current opioid use"
Allocation concealment (selection bias)	Low risk	Allocation concealment was not mentioned in published reports about the trial. Study authors provided the following information regarding allocation concealment method via e‐mail:  "Subjects who signed informed consent received a screening number starting with the letter “S”." Treatment allocation was done on the basis of eligibility of the subject, and subject numbers were assigned at that moment. In other words, after screening, each eligible subject received a subject number in consecutive order, starting with the lowest number available. Randomisation was double‐blind, with dose incrementation. Randomisation was carried out by the pharmacy of UMC St Radboud using MS Excel. The pharmacist delivered the study medication on the subject’s number, without reference to the content. Allocation was present only through this pharmacist, who was contacted again only after closing of the trial master file on the data acquisition and registration in Castor and after the analyses plan was agreed on." Based on this response, we concluded that the central allocation concealment method was used, via a third person who was not involved in other segments of the trial.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Study authors reported that the study was "double‐blind." Furthermore, they reported: "treatment allocation was strictly concealed from participants, caregivers, investigators, and all other personnel directly involved in the study and was not made available until study completion and database lock." Thus, we judged the study to have low risk for blinding of participants and personnel.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Not reported in the published reports. Upon contact with investigators, we received the following information: "all outcome assessors were strictly blinded for the treatment allocation, as none was in contact with the pharmacy on the allocation"
Incomplete outcome data (attrition bias) All outcomes	Low risk	1 of 24 (4%) participants in the intervention group discontinued treatment due to adverse events, and 2 of 26 (8%) participants in the placebo group due to adverse event and withdrawal of consent, respectively. Attrition was small in both groups. Analyses were performed using an intention‐to‐treat method. Data from all subjects were used without imputation.
Selective reporting (reporting bias)	Low risk	The study protocol was registered at www.clinicaltrials.gov (clinical trial identifier number NCT01608217). All outcomes mentioned in the study protocol (NCT01608217) were reported. Results for pain assessment (Pain Assessment Checklist for Seniors with Limited Ability to Communicate Dutch version (PACSLAC‐D) and Verbal Rating Scale (VRS)) are described in Appendix E‐1 and in table E‐1 on the journal website at Neurology.org.
Other bias	Low risk	Other sources of bias not found.

Volicer 1997.

Study characteristics
Methods	Type of RCT: placebo‐controlled cross‐over design Date: not reported Study duration: 12 weeks
Participants	Location: USA Geographical area: USA Inclusion criteria: participants with the diagnosis of "probable" dementia of the Alzheimer type according to DSM‐III‐R and NINCDS‐ADRDA criteria (McKhann 1984), who exhibited simple food refusal Exclusion criteria: not reported Sample: 15 randomised participants Age: 65 + Sex: 14 men, 1 woman Baseline characteristics: patients were hospitalised on the Dementia Study Unit. Duration of dementia of the Alzheimer type (DAT) varied from 2 to 16 years; most patients had been in institutional long‐term care for many months before the start of the study.  Baseline values of the specific tests (mean ± SD) sMMSE: 4.0 ± 7.4   Katz Activity of daily living score: 5.7 ± 0.6  Bedford Alzheimer Nursing Scale ‐ Severity score: 17.5 ± 3.0
Interventions	Intervention: dronabinol Comparator: placebo Crossover study design: the study used a double‐blind placebo‐controlled cross‐over design, with each study period lasting 6 weeks. Dronabinol or placebo was administered on a fixed‐dose schedule, 2.5 mg capsule (Marionol, Roxane Laboratories), or an identical‐looking placebo capsule was given every morning and noon Frequency: 2.5‐mg capsule (Marionol, Roxane Laboratories) or identical‐looking placebo capsule every morning and noon Setting: Department of Veterans Affairs Supervision: patients were identified by a survey of nursing staff. The extent of disturbed behaviour exhibited by each patient was determined each week by interview with primary caregivers who were familiar with patients' behaviour and with rating scales Adherence: nurses gave the therapy on a daily basis to every patient. This administration schedule was a modification of that recommended for AIDS patients because DAT patients usually eat most of their daily food during breakfast and lunch
Outcomes	Body weight Skin fold thickness Caloric intake Behaviour measures: Cohen ‐ Mansfield Agitation Inventory (CMAI) and Lawton Observed Affect Scale ‐ Past Negative and positive affect scores Plasma albumin Lymphocyte count Adverse events
Notes	Funding: the study was supported by the Department of Veterans Affairs, Roxane Laboratories, Inc., and UNIMED Pharmaceuticals, Inc. Declaration of interest: not reported Trial registration: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	It is indicated that participants were randomised, but the method of randomisation was not reported.
Allocation concealment (selection bias)	Unclear risk	Not reported.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Study authors reported only that the study was "double‐blind," but they did not report which individuals exactly were blinded. The term "double‐blind" is used inconsistently, which makes it difficult to know who was blinded.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not reported.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Four participants dropped out; at 10 weeks: one participant in the placebo group who died before the end of the study; in the dronabinol group, one who developed a grand mal seizure after the first dronabinol dose and two who developed serious intercurrent infections. Of the 15 participants included, 4 were lost (27%) and 11 completed both study periods. It is not reported when participants who dropped out due to intercurrent infections terminated the study.
Selective reporting (reporting bias)	Unclear risk	Information regarding the study protocol was not reported.
Other bias	Low risk	Other sources of bias were not found. We consider the washout period to have been adequate for avoidance of carry‐over effects.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Amanullah 2013	Not an RCT (case series)
Aragona 2009	Not a dementia patients sample (multiple sclerosis)
Assogna 2020	Not an RCT
Boxer 2015	Not an RCT; cannabinoids not an intervention (review of trials for disease‐ modifying agents for frontotemporal lobar degeneration)
Broers 2019	Not an RCT (prospective observational pilot study)
Caldentey 2012	Not a dementia patients sample
Chagas 2013	Not a dementia patients sample
Chagas 2014	Not a dementia patients sample
Consroe 1991	Not a dementia patients sample
Curtis 2009	Not a dementia patients sample
Euctr 2016	EudraCT Number: 2016‐004169‐18; status of the trial "prematurely ended"
Ghaffar 2008	Not an RCT; not a dementia patient sample (consecutive study with patients with multiple sclerosis)
Kahraman 2009	Not an RCT (consecutive assessment of patient charts)
Libro 2016	Not an RCT (study with gingival mesenchymal stem cells)
Mahlberg 2007	Randomisation not reported; data not reported separately for melatonin and dronabinol groups
NCT00842985 2008	Not a dementia patients sample (marijuana users)
NCT01964547 2012	Not a dementia patients sample (multiple sclerosis)
Noonan 2010	Not an RCT (study with cultured neurons from rats)
Passmore 2008	Not an RCT (case report)
Scotter 2010	Not an RCT (review of endocannabinoid system in neurodegenerative diseases)
Shelef 2016	Not an RCT (open‐label prospective cohort study)
Solowij 2014	Not a dementia patient sample (healthy volunteers)
Stone 2010	Not a dementia patient sample (healthy volunteers)
Walther 2006	Not an RCT (patients included consecutively)
Walther 2011	This is a failed trial that trial authors discovered they were unable to recruit to. We could classify this as a pair of case reports. The manuscript describes data for 2 participants only
Wilhelm 2017	Not an RCT (case report)
Woodward 2014	Not an RCT (retrospective chart review)
Zajac 2015	Not an RCT (case study)

Characteristics of studies awaiting classification [ordered by study ID]

NCT03328676 2017.

Methods	Type of RCT: phase II, randomised, double‐blind, placebo‐controlled trial Method of allocation: parallel assignment, masking‐triple (participant, care provider, investigator) Date: study start date: December 7, 2017; estimated study completion date: May 2020 Study duration: 3 years (duration of treatment not reported in trial protocol)
Participants	Location: Laniado Hospital, Israel Geographical area: Netanya, Israel Sample: estimated enrolment: 60 participants Age: > 60 years old Sex: all Inclusion criteria: male or female subjects > 60 years old; informed consent from participants' legally authorised representative; subjects residing either in an institutionalised setting (e.g. dementia unit, nursing home, assisted living facility, other residential care facility) or in a non‐institutionalised setting where the subject is not living alone and is receiving 24‐hour supervision via home health care or a family member; subjects must have been at their current location for at least 14 days before screening and must be planning to remain at the same location for the duration of the trial; Diagnosis of Dementia (NCD) according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM‐V) criteria, for at least 6 months before screening, Mini‐Mental State Examination < 23, clinically relevant Behavioral and Psychological Symptoms of Dementia (BPSD) operationally defined as NPI‐NH agitation/aggression subscore ≥ 3 at screening; documented history of clinically relevant BPSD; ability to participate in study evaluation and to ingest oral medication; stable concomitant medications regimen for treatment of BPSD for at least 1 month used before the screening visit; stable concomitant medications regimen for treatment of concurrent conditions for at least 1 month before the screening visit Exclusion criteria: patients receiving any of the following medications: astemizole, cisapride, pimozide, or terfenadine; agitation/aggression attributable to concomitant medications, environmental conditions, or psychiatric condition; severe heart disease, epilepsy, anxiety disorder, alcohol and/or substance abuse; psychiatric condition in the past OR suffering from psychosis; schizophrenia OR family history of schizophrenia OR any other mental disorder or any other condition that in the judgement of the investigator would prevent the subject from completing the study, or any condition that the investigator believes would interfere with the intent of the study or would make participation not in the best interest of the patient; surgery within 30 days before screening or scheduled surgery during the study period
Interventions	Intervention: "Avidekel" cannabis oil Comparator: placebo Intervention delivery: cannabis oil will be made out of extract from the Avidekel strain and olive oil. Avidekel oil containing Δ9‐tetra‐hydrocannabinol (Δ9‐THC) and cannabidiol (CBD) in a 1:20 ratio and at a concentration of 30% CBD and 1.5% Δ9‐THC. Each Avidekel oil drop is approximately 0.04 mL in volume containing about 12 mg CBD and 0.6 mg Δ9‐THC Frequency: patients will receive study medication as drops applied under the tongue 3 times a day ‐ morning, noon, and evening ‐ at a minimum of 4 hours apart between administrations Setting: hospital (Laniado Hospital, Israel)
Outcomes	Primary outcome measure Proportion of subjects achieving a Cohen‐Mansfield Agitation Inventory (CMAI) ≥ 4‐point decrease during the treatment period
Notes	Funding: sponsors and collaborators ‐ TO Pharmaceuticals Trial registration Number of protocol at ClinicalTrials.gov: NCT03328676 Other study ID numbers: TO_D02_2016 Status: the study was completed. Full study report is not yet available; thus the study cannont be assessed presently

Characteristics of ongoing studies [ordered by study ID]

ACTRN12619000474156 2019.

Study name
Methods	Type of RCT: randomised, double‐blind, cross‐over, placebo‐controlled trial Study duration: 16 weeks
Participants	Location: University of Notre Dame, Australia Geographical area: Australis Age: 65 years and older Sex: all Inclusion criteria: living within a residential aged care facility; aged 65 years or older; diagnosis of dementia; ability to speak English; known compliance to taking medication; not bedridden Exclusion criteria: to minimise the likelihood of an adverse event, people with certain health conditions or on some medications will be excluded from the study. These include diagnosis of 1 or more of the following conditions: frontotemporal or Lewy body dementia; other diseases such as epilepsy, anorexia nervosa, comorbid psychiatric condition, Parkinson’s disease, congestive heart failure, history of myocardial infarction or anginal pain, history of stroke, liver disease, renal disease. Taking medications that may interact with cannabis metabolism such as primidone, phenobarbital, carbamazepine, rifampicin, rifabutin, troglitazone, Hypericum perforatum, valproic acid
Interventions	Intervention: medicinal cannabis oil Comparator: placebo Intervention delivery: this trial will include 2 six‐week treatment cycles, with 2 two‐week washout periods. For the first six‐week treatment cycle, participants will be randomised into the placebo group (Arm 1) or the medicinal cannabis oil (Arm 2) group. After the first six‐week treatment cycle, a two‐week washout period will occur. During the second treatment cycle, those who took medicinal cannabis oil (Arm 2) in the first treatment cycle will take placebo (Arm 1) during the second treatment cycle and vice versa. After the second treatment cycle, an additional two‐week washout period will occur to monitor participants, as they will no longer be taking any additional medication. The mode of administration will be an oral spray. The dose will be increased from 2.5 mg/d (1 spray) until the participant reaches the best tolerated dose or a maximum of 50 mg/d (20 sprays). The dose will be increased approximately every 3 days, and on the days the dose has been increased, the resident will record the presence of any adverse events 1 hour after the dose has been administered. Following the recording of any moderate to severe adverse events (determined as ‘Somewhat worse’ (moderate) or ‘Much worse’ (severe) on the participants adverse event record) that have not been ameliorated by the time of the next dose, the participant will receive the previous, best tolerated dose
Outcomes	Primary outcome measure Neuropsychiatric Inventory Questionnaire‐Nursing Homes (NPI‐NH; Cummings et al., 1997) Secondary outcome measures Adverse events, blood pressure, heart rate, weight, body composition (using a scale) to measure lean body mass, bone mass, and fat mass Assessment of primary carer and family member perceptions towards medicinal cannabis Cohen‐Mansfield Agitation Inventory ‐ Community (CMAI‐C; Cohen‐Mansfield, 1996) Quality of life Alzheimer’s disease (QOL‐AD; Logsdon, Gibbons, McCurry, & Teri, 2002) Abbey Pain Assessment Scale (APAS; Abbey et al., 2004)
Starting date
Contact information
Notes	Trial registration Number of protocol at Australia and New Zealand Clinical Trials Register: ACTRN12619000474156 Status: ongoing

Euctr  2010‐024577‐39 2011.

Study name
Methods	Type of RCT: randomised, double‐blind, cross‐over, placebo‐controlled trial
Participants	Included patients must have a completed NPI with minimum score of 10 (or at least 1 item with high frequency and severity) and should have symptoms of agitation or aggression Inclusion criteria: diagnosis of Alzheimer's disease, vascular dementia, or mixed, according to the criteria of NINCDS‐ADRDA/NINCDS‐AIREN; Clinical Dementia Rating score between 0.5 and 2; behavioural problems (especially agression or agitation) related to dementia, lasting for at least 1 month; informed consent from patient and/or from caregiver
Interventions	Intervention: Namisol (delta‐9‐tetrahydorcannabinol) Comparator: placebo
Outcomes	Primary outcome objective and measures Change in total NPI (Neuropsychiatric Inventory) between the 2 consecutive scores per treatment block (score placebo minus score on THC) Secondary outcome objectives To evaluate efficacy on secondary characteristics (e.g. caregiver burden)  To determine the pharmacokinetic profile of THC (tetrahydrocannabinol) and its metabolite 11‐OH‐THC (11‐hydroxy‐delta 9‐THC) and the relationship between plasma concentrations of THC and 11‐OH‐THC and clinical effects in these patients To evaluate the safety and tolerability of THC To evaluate long‐term safety and tolerability in an open‐label extension study To perform concurrent validation of the (Dutch) clinician version of the Neuropsychiatric Inventory (NPI‐C) against the proxy‐based version of the NPI
Starting date
Contact information
Notes	EudraCT Number: 2010‐024577‐39; study status: "ongoing"

Euctr 2019‐002106‐52‐GB.

Study name	Cannabidiol for behavioural symptoms in Alzheimer’s disease (CANBiS‐AD)
Methods	Type of RCT: randomised, double‐blind, placebo‐controlled trial Study duration:  6 weeks
Participants	Location: King's College London, Department of Old Age Psychiatry, London, UK Geographical area: UK Age: age 55 and above Sex: all Inclusion criteria: age 55 and above from both genders; living within the SLAM/South London region; diagnosis of AD according to criteria of National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association (NINCDS‐ADRDA); Clinical Dementia Rating (CDR) score between 0.5 and 2; BPSD lasting at least 2 weeks, with total score on Neuropsychiatric Inventory (NPI) ≥ 4 and at least 1 item with moderate frequency or severity on 1 of the domains of anxiety, agitation, hallucinations or delusions; if treated with cholinesterase inhibitors (donepezil, rivastigmine, or galantamine) and/or memantine, dosage must have been stable for at least 1 month; if ChEIs and/or memantine has been discontinued, may enrol 1 month following discontinuation; ability to participate in study evaluation and to ingest oral medication; reliable informant/caregiver; written informed consent from participant or legally authorised representative Eligibility criteria for caregivers: see the patient for an average of 6 hours per week; willing to attend all study visits with participant and participate in weekly calls; able to answer questions about participant’s behavioural and psychological symptoms, daily activities, quality of life, memory, health, and any side effects experienced from the study medication Exclusion criteria: presence or history of other psychiatric or neurological disorders (e.g. psychotic disorders, schizophrenia, major depressive illness including suicidal ideation, stroke, epilepsy) or severe intercurrent physical illness that could interfere with conduct of the study; participants answering "yes" on the C‐SSRS Suicidal Ideation Item 4 or Item 5 whose most recent episode meeting the criteria for C‐SSRS Item 4 or Item 5 occurred within the last 6 months, OR who answer "yes" on any of the 5 C‐SSRS Suicidal Behavior Items and whose most recent episode meeting the criteria for any of these 5 C‐SSRS Suicidal Behavior iItems occurred within the last 2 years, OR who, in the opinion of the investigator, presents a serious risk of suicide; changes in dosage of antidepressants within 4 weeks before randomisation and during study, and changes in dosage of antipsychotics or benzodiazepines within 2 weeks before randomisation and during study; use of co‐medication that has a clinically relevant interaction with CYP2C19 or CYP3A classes of liver enzymes will not be permitted from 2 weeks before inclusion until end of the study. Examples of co‐medication that will be not allowed include CYP3A4 inhibitors (such as itraconazole, ketoconazole, posaconazole, fluconazole, erythromycin, clarithromycin, telithromycin, HIV protease inhibitors, nefazodone, telaprevir, boceprevir, imatinib, ticagrelor, voriconazole), CYP3A4 inducers (such as grapefruit juice, carbamazepine, efavirenz, nevirapin, etravirin), and CYP2C19 inhibitors (such as moclobemine, fluvoxamine, chloramphenicol, fluoxetine); if the patient has participated in other interventional clinical studies within 30 days of baseline; female patients who are pregnant or lactating; female patients of childbearing potential* who are not willing to use a highly effective method of contraception** for the duration of the trial (from date of consent until entire study duration, i.e. until the 4‐week follow‐up visit post treatment) to prevent pregnancy or abstain from heterosexual activity. *Females of childbearing potential are females who have experienced menarche and are not surgically sterilised (e.g. hysterectomy, bilateral salpingectomy) or postmenopausal (defined as at least 1 year since last regular menstrual period); known hypersensitivity to CBD gelatin or micro‐crystalline cellulose; mechanistic sub‐study only: any contraindication to MRI, including pacemakers, metallic foreign body in the eye, aneurysm clip in the brain, severe claustrophobia whereby patient would not be able to tolerate the scan, etc
Interventions	Intervention: cannabidiol, capsule, concentration range 200 to 600 mg Comparator: placebo, capsule
Outcomes	Primary outcomes  Acceptability: to investigate acceptability of the randomisation procedure and delivery of the intervention (CBD or placebo in capsule form) to AD patients and carers (recruitment rates) Number of participants recruited, consented, and randomised in 9 months Retention: to measure the study retention rate (follow‐up rates) Measured by dropout by the end of 6 weeks of treatment and missing data Compliance: to assess compliance with treatment (adherence) Percentage of patients compliant with treatment by the end of 6 weeks (compliance defined as at least 60% of the dose taken, measured by pill count) Secondary outcomes  To measure key outcome domains for patients including clinical measures ‐ BPSD, cognition, quality of life, smell functioning, daily functioning ‐ and carergivers' distress with selected scales (NeuroPsychiatry Inventory (NPI‐C); Standardised Mini‐Mental State Examination (sMMSE); Addenbrooke's Cognitive Examination‐III (ACE‐III); Hopkins Verbal Learning Test (HVLT); Cohen Mansfield Agitation Inventory (CMAI); Clinical Global Impression of Change (CGIC); DEMQOL and DEMQOL‐Proxy; Bristol Activities of Daily Living Scale (BADLS); Smell Functioning Test (UPSIT)) To assess the safety (adverse events) and tolerability (side effects) of CBD treatment (self‐report of side effects, medication discontinuation, and UKU side‐effect rating scale for psychotropic drugs); safety parameters: bloods ‐ haematology (haemoglobin, red blood cell count, platelet count, white blood cell count, differential count); clinical chemistry (sodium, potassium, creatinine, urea); glucose; liver function test (total bilirubin, alanine aminotransferase, aspartate aminotransferase, alkaline phosphatase); cholesterol (total, LDL, HDL); T3, T4, and thyroid‐stimulating hormone; vital signs: heart rate and blood pressure; physical examination; electrocardiogram (ECG) ‐ 12‐lead ECG) To describe the variance and distribution of quantitative outcomes to pilot effect sizes and inform design and sample size calculation in the future large‐scale RCT To investigate acceptability of the questionnaires used to measure key outcome domains (percentage of completion rates) Additional outcomes Mechanistic sub‐study: to investigate the mechanistic basis of any CBD effect observed during RCT using functional brain imaging and spectroscopy Qualitative interviews: to explore participants' experiences of taking part in the trial including what encouraged them to take part, their views on the intervention, and how they are finding the trial in general
Starting date	2020‐03‐19
Contact information	King's College London; Dr. Latha Velayudhan; latha.velayudhan@kcl.ac.uk
Notes	EudraCT number: 2019‐002106‐52 Status: ongoing

Euctr 2020‐001056‐17‐GB.

Study name	A randomised feasibility trial investigating Sativex® for the treatment of the agitation and aggression (A/A) in Alzheimer’s dementia (STAND)
Methods	Type of RCT: randomised, double‐blind, parallel‐design, placebo‐controlled trial Study duration: 8 weeks
Participants	Location: Institute of Psychiatry, Psychology & Neuroscience (loPPN), London, UK Geographical area: UK Age: 55 to 90 years Sex: all Inclusion criteria: probable Alzheimer’s disease diagnosis according to criteria of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association (NINCDSADRDA); clinically significant A/A that requires treatment, defined by CMAI >/= 45 and/or NPI‐NH Agitation Total Score >/= 4; residential within a nursing home at recruitment to the study with a history of at least 2 weeks behavioural disturbance; written and witnessed informed consent from participant (if deemed having mental capacity) or from personal legal representative (next of kin/power of attorney) or from professional legal representative (non‐R/F member who can attest to knowing prospective participant for significant period of time) Exclusion criteria: anti‐psychotic, anti‐epileptic, antidepressant, benzodiazepine, lithium or hypnotic dosage alteration in the 2 weeks before the start of the study (must be expected to maintain dosage throughout the study); ChEIs (donepezil, rivastigmine, or galantamine) and/or memantine, dosage alteration in the 6 weeks before the start of the study; currently using cannabis‐based medicine(s) (defined as a UK licensed product prescribed by a doctor); concomitant treatment with strong enzyme inducers (rifampicin, carbamazepine, phenytoin, phenobarbital, St John’s Wort) and/or CYP3A4 inhibitors; hypersensitivity to Sativex® or any of the excipients in the formulation (ethanol anhydrous, propylene glycol, peppermint oil); severe cardiovascular disease, recent myocardial infarction (‘recency’ determined by study doctor according to clinical significance), uncompensated congestive heart failure, and uncontrolled hypertension; QT interval by Fredericia (QTcF) greater than 450 will be excluded if ECG conducted; severe, unstable, or poorly controlled medical illness; renal Impairment as defined by estimated glomerular filtration rate (eGFR) less than 45 mL/min; hepatic impairment as defined by alanine aminotransferase (ALT)/aspartate aminotransferase (AST) levels 3 times greater than reference value of laboratory (165 IU/L+ for ALT; 150 IU/L+ for AST); any disability that may interfere with the patient completing the study procedure; history or family history of schizophrenia or other psychotic illness; history of severe personality disorder or other significant psychiatric disorder other than depression associated with underlying condition; delirium, pain, or any medical illness as a clear cause of agitation; females of childbearing potential, defined as ‘having experienced menarche and not permanently sterilised (e.g. by hysterectomy, bilateral salpingectomy, bilateral oophorectomy) or postmenopausal (defined as at least 1 year since last regular menstrual period); evidence of ‘suicidality risk’ determined by > 0 on Columbia‐Suicide Severity Rating Scale (C‐SSRS); history/current seizure disorder; history/current alcohol or other substance abuse; history of fall(s) within the last 6 months; COVID‐19‐specific exclusion criteria (positive COVID test result within previous 4 weeks; currently experiencing CV19 symptoms (continuous cough, high temperature, loss of taste/smell). NB: can be re‐screened after 2 weeks
Interventions	Intervention: Sativex Oromucosal Spray (DELTA‐9‐TETRAHYDROCANNABINOL) Comparator: placebo
Outcomes	Primary outcomes  Feasibility To explore rates of recruitment and retention in the target population, including determining facilitators and barriers To determine whether the cutoff of clinically significant agitation for CMAI or NPI‐NH influences rate of recruitment for a future confirmatory trial To investigate acceptability of an oral mucosal method of administration for this indication in terms of compliance and for care home staff in terms of adherence to the titration schedule To investigate acceptability of a cannabinoid‐based medicine, and to explore impact of societal attitudes and stigma within this patient population as part of the qualitative evaluation  Secondary outcomes To estimate the parameters in a sample size for a later RCT for the treatment effect of Sativex® for agitation and aggression in dementia To explore the overall response rate of Sativex® in the trial population, including other neuropsychiatric symptoms, such as sleep disturbances and changes in appetite, and pain relief To explore actigraphy data for sleep, physical activity, and linked to NPS assessments (exploratory analyses) To assess safety and tolerability: assessed by self‐ and carer report of side effects and medication discontinuation and safety parameters (bloods for haematology (full blood count and haemoglobin) and biochemistry (urea and electrolytes, liver function test, thyroid function test, lipid profile)); vital signs: heart rate and blood pressure; physical examination; follow‐up phone calls for self‐reported side effects, including incidence of falls and compliance Tertiary outcomes Changes in neuropsychiatric symptoms, cognition, pain, quality of life measured with selected scales (CMAI,  Neuropsychiatric Inventory ‐ Nursing Home (NPI‐NH), Mini Mental State Examination (MMSE),  Functional Assessment Staging of Alzheimer’s Disease (FAST), Clinical Frailty Scale (CFS), Quality of Life with resident (QOL‐AD care home), Quality of Life with proxy informant 4 and 8 weeks, Abbey Pain Scale (APS)), adverse events, serious adverse events, physical activity, sedentary behaviour and sleep disturbances from raw accelerometer data, co‐prescribed psychotropic medications, number of recorded incidents of aggression, number of occasions rescue treatment utilised
Starting date	2020‐08‐14
Contact information	King's College London; Chris Albertyn; chris.albertyn@kcl.ac.uk
Notes	EudraCT Number: 2020‐001056‐17 Status: ongoing

Forester 2017.

Study name
Methods	Type of RCT: randomised, double‐blind, placebo‐controlled trial Method of allocation: parallel assignment, masking‐quadruple (participant, care provider, investigator, outcomes assessor) Date: study start date: March 1, 2017; estimated study completion date: May 2022 Study duration: 3 years (duration of tretment: 3 weeks)
Participants	Location: Johns Hopkins University, Maryland, USA, and Mclean Hospital, Massachusetts, USA Geographical area: USA (Maryland and Massachusetts) Sample: estimated enrolment‐ 160 (Clinicaltrials.gov) or 80 (abstract Forester 2017) Age: 60 to 90 years of age Sex: all Inclusion criteria: diagnosis of dementia due to AD; presence of Agit‐AD as defined by provisional criteria from the International Psychogeriatric Association (IPA); clinically significant severity of agitation defined by NPI‐C Agitation or NPI‐C Aggression > 4; ability to give informed consent, or deemed to lack such capacity by clinical team and legally authorised representative consents; fluency in English and/or Spanish (includes reading, writing, and speech); ablity to stay at McLean Hospital, Miami Jewish Health, or Johns Hopkins Hospital, for the study duration (3 weeks); 60 to 90 years old; beginning enrolment in study within 1 week of being determined eligible Exclusion criteria: serious or unstable medical illness that might confound assessment of safety outcomes (cardiovascular, hepatic, renal, respiratory, endocrine, neurologic, or haematologic disease); seizure disorder; baseline delirium as determined by Confusion Assessment Method (CAM) and Diagnostic and Statistical Manual of Mental Disorders (DSM)‐5 criteria; current use of lithium; inability to swallow a pill
Interventions	Intervention: dronabinol (Marinol) Comparator: placebo Intervention delivery: capsules of dronabinol will contain 2.5 mg per dose (5 mg daily) during Week 1, then increased to 5 mg per dose (10 mg daily) for Weeks 2 and 3 Frequency: study medication will be administered twice daily Setting: hospital (McLean Hospital, Miami Jewish Health, or Johns Hopkins Hospital)
Outcomes	In both protocol and abstract: Primary outcome measure Symptoms of agitation: Pittsburgh Agitation Scale (PAS) and the Neuropsychiatric Inventory, Clinical version (NPI‐C) Secondary outcome measures Adverse events (AEs) In abstract (Forester 2017) Besides PAS, NPI‐C, and AEs, diagram shows Cohen‐Mansfield Agitation Inventory (CMAI), Clinical Global Impression of Change (CGI‐C), Confusion Assessment Method (CAM)
Starting date
Contact information
Notes	Funding: sponsors and collaborators ‐ Johns Hopkins University, Mclean Hospital Trial registration Number of protocol at ClinicalTrials.gov: NCT02792257 Other study ID numbers: IRB00052955 Status: recruiting

NCT04436081 2020.

Study name	Effects of THC‐free CBD oil on agitation in patients with Alzheimer's disease
Methods	Type of RCT: randomised, double‐blinded, placebo‐controlled, cross‐over trial  Study duration: 6 weeks
Participants	Location: Eastern Virginia Medical School ‐ Recruiting ‐ Norfolk, Virginia, United States Geographical area: USA Age: 50 to 90 years Sex: all Inclusion criteria: males/females over 50 years old; diagnosis of dementia due to AD or mixed AD with another type of dementia; Mini‐Mental State Exam score (MMSE) between 4 and 28 inclusive; presence of agitation with a Neuropsychiatric Inventory (NPI)‐agitation/aggression subscore > 3; participants and informal caregivers must be fluent in English (includes reading, writing, and speech) and able to give informed consent; for patients treated with cognitive‐enhancing medications (cholinesterase inhibitors (ChEIs) and/or memantine), dosage must be stable for at least 1 month (30 days). If the ChEI and/or memantine has been discontinued, patients may enrol after 15 days; eligible caregivers must live with the participant or have a minimum of 4 hours of daily contact with the participant Exclusion criteria: diagnosis of non‐AD or non‐mixed dementias; very mild dementia or advanced dementia (MMSE: greater than 28 or less than 4); NPI‐agitation‐aggression score < 3; serious or unstable medical illness including cardiovascular, hepatic, renal, respiratory, endocrine, neurologic, or hematologic disease that might confound assessment of safety outcomes as determined by study physician; presence or history of other serious psychiatric disorder or neurologic condition (e.g. psychotic disorder, bipolar disorder, schizophrenia); current abuse of/dependence on marijuana; current drug abuse, current alcohol abuse; seizure disorder; pregnant or breastfeeding; indication of baseline delirium as determined by the Confusion Assessment Method (CAM); current use of lithium; inability to swallow CBD oil softgels; change in dosage of anti‐depressive within 4 weeks before randomisation and during the study; change in dosage of antipsychotics or benzodiazepines within 1 week before randomisation and during the study; contraindication to CBD oil (history of hypersensitivity to any cannabinoid); frequent falling due to orthostatic hypotension; use of tricyclic antidepressants (TCAs), fluoxetine, and/or carbamazepine. Residing in a nursing home
Interventions	Intervention: hemp‐based CBD oil gelcaps Comparator: oral placebo gelcaps Intervention delivery: intervention consists of 6 weeks' oral administration of CBD oil gelcaps, starting at a dosage of 15 mg twice per day with up‐titration to 45 mg twice per day. In case of side effects, current dosage will be reduced to the previous one. Participants n the placebo group will receive oral placebo gelcaps that are identical in appearance to CBD oil gelcaps, with identical dosing as in the CBD oil gelcaps group
Outcomes	Primary outcomes Change in agitation and aggression  measured by the Cohen‐Mansfield Agitation Inventory (CMAI), a validated 29‐item questionnaire to assess agitation Change in caregiver burden measured by the Zarit Burden Interview (ZBI), a validated 22‐item questionnaire to assess caregiver burden Change in participant's quality of life measured by the Quality‐of‐Life Assessment in Dementia (DEMQOL‐proxy) Change in caregiver's quality of life measured by the Measurement of Quality of Life in Family Carers of People With Dementia (C‐DEMQOL) Secondary outcomes Assessment of change in neuropsychiatric symptoms measured by the Neuropsychiatric Inventory (NPI) Assessment of change in cognitive skills measured by the Mini Mental State Exam (MMSE), a validated 30‐item questionnaire used to measure cognitive impairment among the elderly Effects of CBD oil on sleep quantity measured by the actigraphy function of fit bit
Starting date	February 26, 2021
Contact information	Patricia Sandoval; sandovpm@evms.edu
Notes	Number of protocol on ClinicalTrials.gov: NCT04436081 Status: recruiting

NCT04516057 2020.

Study name	Nabilone for agitation blinded intervention trial (NAB‐IT)
Methods	Type of RCT: randomised, double‐blind, parallel‐design, placebo‐controlled trial Study duration:  8 weeks' treatment + 8 weeks' follow‐up
Participants	Location: Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada Geographical area: Canada Age: age 55 and above Sex: all Inclusion criteria: males or females ≥ 55 years of age; postmenopausal females; Diagnostic and Statistical Manual of Mental Disorders‐5 (DSM 5) criteria for Major Neurocognitive Disorder due to AD (patients with major neurocognitive disorder due to multiple etiologies (AD and vascular)); sMMSE ≤ 24; presence of clinically significant agitation based on the IPA definition; If treated with cognitive‐enhancing medications (cholinesterase inhibitors and/or memantine), dosage must be stable for at least 3 months; availability of a primary caregiver to accompany the participant to study visits and to participate in the study. Primary caregiver must be sufficiently proficient in English to complete required study assessments, as per investigator judgement Exclusion criteria: change in psychotropic medications less than 1 week before study randomisation (e.g. concomitant antidepressants); contraindication to cannabinoids (e.g. allergies to cannabis and cannabis products), potential clinically important drug‐drug interactions; current uncontrolled cardiovascular disease (e.g. uncontrolled hypertension, ischaemic heart disease, arrhythmia, severe heart failure), as per investigator assessment; current significant liver disease, as per investigator assessment; presence or history of other psychiatric disorders or neurological conditions (e.g. psychotic disorders, schizophrenia, stroke, epilepsy); currently meeting DSM‐5 criteria for major depressive episode (MDE); previous or current abuse of/dependence on marijuana; clinically significant delusions and/or hallucinations (NPI‐NH delusion/hallucinations subscore ≥ 4); current reported recreational use of marijuana or other cannabis products
Interventions	Intervention: nabilone Comparator: placebo Intervention delivery: in the nabilone arm, participants will receive nabilone titrated up to a maximum dose of 2 mg/d. Pariticpants in the placebo arm will receive placebo capsules
Outcomes	Primary outcome: Agitation measured with the Cohen‐Mansfield Agitation Inventory (CMAI) Secondary outcomes Behaviour measured with the Neuropsychiatric Inventory ‐ Nursing Home (NPI‐NH) [Time Frame: Baseline (0 Weeks) to 8 Weeks] ‐ a widely used assessment of behaviour disturbances in dementia, including apathy, agitation, delusions, hallucinations, depression, euphoria, aberrant motor behaviour, irritability, disinhibition, anxiety, sleeping, and eating. Frequency and severity of these symptoms are judged on a 4‐point and 3‐point scale, respectively. Scores range from 0 to 144 points, with higher score indicating worse outcome Cognition measured with Standardised Mini‐Mental State Examination (sMMSE)  Weight  Nutritional status measured with Mini Nutritional Assessment ‐ Short Form (MNA‐SF)  Pain measured with the Pain Assessment Checklist for Seniors with Limited Ability to Communicate‐II (PACSLAC‐II) Global change ‐ Alzheimer's disease cooperative study ‐ measured with Clinical Global Impression of Severity/Change (ADCS‐CGIS/C) Other Outcome Measures Sedation measured with Sleepiness/Sedation subscale of the UKU‐Side Effect Rating Scale
Starting date	February 1, 2021
Contact information	NAB‐IT Coordinating Centre; NAB‐IT@sunnybrook.ca
Notes	Number of protocol on ClinicalTrials.gov: NCT04516057 Status: recruiting

Differences between protocol and review

Differences made in the review protocol from the first version of the review/protocol

The original protocol for this review was published in 2008, and the first version of the review was published in 2009 (Krishnan 2009). The following changes to the review methods compared to the original protocol were made to bring them in line with current Cochrane recommendations:

1. Diagnostic criteria

The protocol now specifies updated references for diagnostic criteria that will be used for dementia.

2. Type of studies

Any RCTs on the subject will be included. Any comparator will be considered.

3. Primary and secondary outcomes

The new version of the protocol updated primary and secondary outcomes, specifying that validated measures for primary outcomes will be analysed, and that biomarker outcomes will not be included.

4. Search methods for identification of studies

This section was updated to include information about ALOIS, the CD‐CIG Specialised Register.

5. Selection of studies

All references will be managed using EndNote software. A process for screening studies and resolving disagreements has been now described in detail.

6. Data extraction and management

This section was rewritten extensively. It now indicates that data extraction from figures will be done by contacting the authors of included studies or via open‐source software Plot Digitizer.

7. Assessment of risk of bias in included studies

This section was added. Previous section “Quality assessment” was removed. Use of the Cochrane RoB tool was described.

8. The following new sub‐sections were added

Measures of treatment effect, Unit of analysis issues, Dealing with missing data, Assessment of heterogeneity, Assessment of reporting biases, Data synthesis, Subgroup analysis and investigation of heterogeneity, Sensitivity analysis, Presentation of results – ‘Summary of findings’ tables.

Relevant explanations were written for all these sub‐sections.

Differences between the protocol published in 2017 and the review

In the protocol published in 2017, we planned the following: "Since dementia is a progressive disease, we will include only the data of the first period of cross‐over RCTs." However, our search for evidence revealed multiple cross‐over trials (Herrmann 2019; van den Elsen NCT01302340;  Volicer 1997), which did not report many data separately for each study period and were of short duration (12 to 14 weeks). We included those cross‐over studies, as we did not have good reason to exclude short cross‐over trials targeting behavioural and psychosocial symptoms of dementia (BPSD), especially when correctly analysed (paired data).

In the protocol published in 2017, we stated that we planned to conduct random‐effects models of meta‐analyses. However, in the review, we used a fixed‐effect model. The random‐effects model relies on the estimation of between‐study heterogeneity. When only a small number of studies are available, such estimation is often inaccurate and results in biased effect estimates and too narrow confidence intervals. In such cases, the fixed‐effect model is considered as an appropriate model. Given that the statistician AJ joined our team after submission of protocol, the effect model was introduced into analysis at the level of the full review.

Contributions of authors

All authors contributed to searching the relevant literature, providing intellectual content, and writing the updated version of the protocol. All authors approved the final version of the protocol.

DBK, DM, TB, MJK, ZR, and AVV conducted screening and data extraction and contributed to writing of the review. DBK additionally performed GRADE analysis.

AJ verified data extraction, performed data analysis, interpreted meta‐analyses, and contributed to writing of the review.

LP coordinated the review process, arbitrated screening and data extraction decisions, performed GRADE analysis, and contributed to writing of the review.

Sources of support

Internal sources

• No sources of support provided

External sources

• NIHR, UK

  This review was supported by the National Institute for Health Research (NIHR), via Cochrane Infrastructure funding to the Cochrane Dementia and Cognitive Improvement Group. The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the Systematic Reviews Programme, NIHR, National Health Service, or the Department of Health

Declarations of interest

The review authors have no conflict of interest to declare.

New