Volicer 1997.

Methods	Randomised double blind placebo‐controlled cross‐over trial (randomization method not given)
Participants	Country: USA Single centre Subjects: 15 patients hospitalised in a Dementia Study Unit Selection criteria: diagnosis of probable Alzheimer's dementia with simple food refusal, normal blood tests or no significant laboratory abnormalities Exclusion criteria: problems with choking on food and liquids, hypersensitivity to dronabinol or sesame oil. Age range: 65‐82 years, 11 male, 1 female (information only given for 12 patients)
Interventions	Placebo or Dronabinol 2.5 mg capsule twice daily for 6 weeks followed by switch to the alternative treatment for a further 6 weeks
Outcomes	Weight gain, body mass index (BMI), triceps skin fold thickness, caloric intake Disturbed behaviour (Cohen‐Mansfield Agitation Inventory score) Affect (Lawton Observed Affect Scale ‐ Past) Plasma albumin and lymphocyte count
Notes	Small study group. Short duration of treatment. Almost all participants in the analysis were male.
Risk of bias
Bias	Authors' judgement	Support for judgement
Adequate sequence generation?	Unclear risk	Quote: "subjects were randomly assigned to placebo first or dronabinol first groups". Comment: No further information provided.
Allocation concealment?	Unclear risk	B ‐ Unclear Comment: No information provided.
Blinding? Caregiver‐reported outcomes	Unclear risk	Quote: "the study used a double‐blind...design"; "Although the study was double‐blind, the staff knew the objectives of the study." Comment: Measurement of disturbed behaviour was based on caregiver interviews but it was not specifically stated that both caregivers and outcome assessors were blinded.
Blinding? Body weight	Low risk	Quote: "the study used a double‐blind...design" Comment: Probably done if measured by outcome assessors.
Incomplete outcome data addressed? All outcomes	High risk	4/15 participants did not complete the study; three of these were omitted from the analysis.
Free of selective reporting?	Low risk	All outcomes were reported on although there was a lack of quantitative data overall.
Free of other bias?	High risk	No wash‐out period leading to risk of carry‐over effect. Pharmaceutical company support could lead to reporting bias.