There are a number of strengths of our study. First, all participants will have a medical diagnosis of dementia. This ensures that the diagnosis is in line with the criteria of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition . Second, the residential aged-care staff spend a large amount of time with the participants, so they will be able to observe small changes, leading to accuracy in recording of the results. When possible, these results will also be compared with information from the participants and their next of kin to examine similarities and differences. Beattie et al.  published a protocol paper outlining a national project to collect multiple QOL perspectives from the care staff, family members and those living with dementia. Comparisons between care staff, family members and self-report QOL scores showed a linear relationship between reporters, with the residents often rating their QOL higher than the care staff [57, 58].
The results of questionnaires completed on behalf of the aged-care staff, participants and family members will be analysed using IBM SPSS Statistics version 25 software (IBM, Armonk, NY, USA). The responses from the aged-care staff will be the main responses considered for analysis. Where available, participants and family responses will be included for secondary analysis. To examine group differences, the participants will be categorised according to their treatment cycle group allocation (group A or group B). Descriptive statistics will be derived. Each variable will be tested for normality. For those variables that meet the normality assumption, two-sided paired and/or independent t tests will be used to examine group differences within and between groups. If the normality assumption is violated, then non-parametric tests such as the Wilcoxon signed-rank test will be used. Within-subject differences of the four measurements between the first and second washout periods will be tested using paired t tests to ensure that the washout phase is long enough to rule out any carryover effects [51, 52]. All data collection points will be examined using general linear mixed modelling techniques to see if any changes in behaviour, QOL or pain have occurred over the duration of both treatment cycles. The covariates of weight, average dose of medication, and baseline measures will be controlled for in each model, and any interactions will be tested and reported. The proportion of adverse events during the CBM and placebo phases will be tested and reported for each individual. NPI-NH is the primary outcome measure for this study. All other measures (CMAI, QOL-AD, and Abbey Pain Scale) have been included for secondary analysis. The CMAI will be analysed using the reliability change methodology compared with the NPI-NH to allow small changes to be reported . The α-value will be set at 0.05. In the instance where a participant withdraws halfway through a treatment cycle, the information collected prior to withdrawal will be retained in the study because their personal information will have been de-identified. The data management of the information collected will follow standard university procedures, be stored in a locked cabinet for a period of 15 years, be stored on a password-protected computer, and be backed up regularly in a secure format.
Cannabinoid-based medicines (CBMs) have been shown to improve dementia symptoms such as aggression and agitation [12, 13], and they appear to be safer to prescribe than other pharmacotherapies  because the adverse effects are often minimal . For example, Weier and Hall  found sedation to be the only adverse effect among patients with dementia prescribed either cannabinoids or pharmacotherapies. While periods of euphoria, somnolence and tiredness were observed among those prescribed dronabinol, a synthetically derived delta-9-tetrahydrocannabinol (THC) , there were only a small number of adverse events (6 of 98) related to the administration of the synthetic THC similar to those manifested by the placebo [14, 17]. However, well-designed, randomised, double-blind, placebo-controlled trials need to be completed to understand the most efficacious formulation, safety profile, drug–drug interactions and true effect to determine the place of CBM in dementia [18, 19], allowing greater generalisability of these outcomes .
This study has received approval from the Human Ethics Research Committee at the University of Notre Dame Australia. The study will use a parallel mixed methods design. The research methodology for this study is a phase II, randomised, placebo-controlled crossover trial. The design will include a 2-week eligibility (assessment) period, two 6-week-treatment cycles to allow each participant to take part in both the control and treatment cycles, and two 2-week washout periods (one between both treatment cycles and the other after the second treatment cycle). The 6-week treatment cycles have been selected on the basis of safety, pharmacodynamics, and pharmacokinetics as reported by Ahmed et al. , and a 2-week washout period has been shown to be safe and an appropriate length of time for cannabis to metabolise in older individuals .
Dementia is the second leading burden of disability among Australians aged 65 years and older, and the burden of disease is expected to increase exponentially over the next 30 years . Alzheimer’s disease is the most common cause of dementia and affects approximately 50–70% of the elderly with dementia. Pharmacological management of behavioural and physical symptoms of dementia is currently the most common treatment option, and many are prescribed medications such as off-label antipsychotics, sedative/hypnotics, anxiolytics, acetylcholinesterase inhibitors, and antidepressants to mask and alleviate the array of dementia symptoms [6, 7]. Medications such as aripiprazole, olanzapine, risperidone and memantine have been shown to reduce troublesome behaviours , although unclear guidelines are often provided for administration [2, 7]. This results in polypharmacy and its inherent risks, with numerous medications being prescribed for a longer duration than recommended . Many of these medications lead to a number of substantial side effects , including increased rate of stroke and mortality .
This phase of the study will take 16 weeks to complete. To minimise the risk of adverse events and variation in the maximum tolerated dose of CBM oil, each participant will receive one dose on the first and second days (2 pm) and two doses (9 am and 2 pm) for the reminder of both treatment cycles. A registered nurse will administer the dose along with a small meal (e.g., morning and afternoon tea), and the rate of titration will be monitored by the pharmacist to ensure it is appropriate for each individual. The participant will gradually receive an increased dose (titration) of the medication over several weeks, as shown in Table 1. During these weeks, the participant along with the care staff will record the presence of, and any change in, any potential adverse events that may be associated with the medication after the first dose, each afternoon when the dose is increased, and again on the final day of medication. If an adverse event is noted, the participant will revert to their previous best tolerated dose using the adverse events and safety protocol listed below.
To our knowledge, this is one of the first trials within Australia to evaluate the use of a purified CBM oil at the individual level among those with dementia to examine behavioural effects, QOL, and pain and discomfort. Only a handful of crossover trials have been conducted [14, 16, 41, 42, 54], although the majority have used fixed doses and have not incorporated an individually tailored dosing regimen. Soto et al.  reviewed 18 randomised clinical trials examining drugs prescribed for agitation and aggression and found large variations not only in the chosen questionnaires to measure these symptoms but also in the inclusion criteria. On the basis of their results, Soto et al.  suggested that trials lasting 9–12 weeks were adequate for assessing an acute response, whereas longer trials (6–12 months) were effective for assessing the stability of a response. The 18-week duration of this study is appropriate to assess the initial dose response , and the trial design is also reflective of other protocols of crossover studies and includes two 2-week washout periods to ensure patient safety and a chance for the medication to metabolise out of the body. For example, Babalonis et al.  designed a protocol to examine the use of a THC:CBD oromucosal spray among post-stroke spasticity patients and included two 4-week treatment cycles with a 2-week washout period between both cycles.
The randomisation process for this study will be done by creating a random number list using a 1:1 ratio allocation to ensure an equal number of cases in both the placebo group (n = 25) and the treatment group (n = 25) using Excel software (Microsoft Corp., Redmond, WA, USA). The determination of participant allocation will be completed by the laboratory manager in the drug manufacturing laboratory, with each case being provided a unique identification (ID) number (1–50). The primary researcher, who is responsible for recruitment, will provide the laboratory manager with the participant’s name to be sequentially matched against with the next available ID number. The laboratory will provide the pharmacy with both CBM and placebo in identical bottles labelled with the ID, but the medical practitioner and research team members will not know the order of treatment until the completion of the study. The pharmacist will place the participant’s name on the bottle before distributing the bottles to the aged-care facilities.
Individuals who express interest in participating in the study will initially be screened on the basis of inclusion criteria (described above). Following the initial screening process, potential participants will undergo a thorough clinical investigation by a geriatrician to ensure they have the cognitive capacity to provide informed consent using the MMSE. The MMSE  is the most widely used cognitive outcome measure to assess the severity of cognitive performance. It comprises 11 items, where a total score out of 30 can be calculated to assess the severity of dementia (25–30 = questionably significant, 20–25 = mild, 10–20 = moderate, 0–10 = severe). Those who seem suitable will be revisited by the geriatrician 1 week after the cognitive tests and will confirm that the participant has understood the purpose of the trial and has recalled the details of the study. Then the primary researcher will invite the eligible participants to enrol in the study and ask them to complete the consent form and provide some demographic and baseline information, including age, sex, education level, weight, medical history including comorbid illnesses, and prescribed medications. The participants will then be randomly allocated to treatment group A or B and receive either CBM oil or placebo for the first 6-week treatment cycle. No adjustments will be made to the participants’ currently prescribed medications.
Brain cells have switches known as receptors that can be activated by endocannabinoids, a class of lipid molecules made by the body that are used for intercellular signaling in the brain. The psychoactive effects of marijuana are caused by THC, a molecule similar in activity to endocannabinoids that can activate the same receptors. Physical activity results in the production of endocannabinoids and some studies have shown that exercise may slow the progression of Alzheimer’s disease.
“Although other studies have offered evidence that cannabinoids might be neuroprotective against the symptoms of Alzheimer’s, we believe our study is the first to demonstrate that cannabinoids affect both inflammation and amyloid beta accumulation in nerve cells,” says Salk Professor David Schubert, the senior author of the paper.
“Inflammation within the brain is a major component of the damage associated with Alzheimer’s disease, but it has always been assumed that this response was coming from immune-like cells in the brain, not the nerve cells themselves,” says Antonio Currais, a postdoctoral researcher in Schubert’s laboratory and first author of the paper. “When we were able to identify the molecular basis of the inflammatory response to amyloid beta, it became clear that THC-like compounds that the nerve cells make themselves may be involved in protecting the cells from dying.”
Cannabinoids remove plaque-forming Alzheimer’s proteins from brain cells
It has long been known that amyloid beta accumulates within the nerve cells of the aging brain well before the appearance of Alzheimer’s disease symptoms and plaques. Amyloid beta is a major component of the plaque deposits that are a hallmark of the disease. But the precise role of amyloid beta and the plaques it forms in the disease process remains unclear.
Other authors on the paper include Oswald Quehenberger and Aaron Armando at the University of California, San Diego; and Pamela Maher and Daniel Daughtery at the Salk Institute.
The researchers found that high levels of amyloid beta were associated with cellular inflammation and higher rates of neuron death. They demonstrated that exposing the cells to THC reduced amyloid beta protein levels and eliminated the inflammatory response from the nerve cells caused by the protein, thereby allowing the nerve cells to survive.
In a manuscript published in June 2016’s Aging and Mechanisms of Disease, the Salk team studied nerve cells altered to produce high levels of amyloid beta to mimic aspects of Alzheimer’s disease.
In separate but related research, his lab found an Alzheimer’s drug candidate called J147 that also removes amyloid beta from nerve cells and reduces the inflammatory response in both nerve cells and the brain. It was the study of J147 that led the scientists to discover that endocannabinoids are involved in the removal of amyloid beta and the reduction of inflammation.