Devanand 1998.

Study characteristics
Methods	Study design: randomised controlled trial Study grouping: cross‐over Study duration: 6 weeks Rescue medication: not allowed for NPS or EPS.
Participants	Number randomised: 66 (42 haloperidol, 24 placebo) Mean age: 72.1 years Sex (female): 64.8% Type of dementia: Alzheimer's disease Severity of dementia: moderate‐severe Indication: diverse NPS, but the majority showed psychosis (71.8%), and psychomotor agitation (78.9%) (baseline psychosis BPRS: 6.8, psychomotor agitation BSS: 3.6) Setting: outpatients Country: USA
Interventions	Intervention characteristics Haloperidol dose:low dose of 0.50 mg to 0.75 mg/day; or standard dose of 2 mg to 3 mg/day. Placebo After the first week the daily dose was raised from two capsules (haloperidol, 2 mg or 0.50 mg, or placebo) to three capsules (haloperidol, 3mg or 0.75 mg, or placebo). If side effects (e.g. extrapyramidal signs) were limiting, on the basis of the psychiatrist’s clinical judgment, the dose was maintained at two capsules daily. Therefore, in both phases, patients were on a stable dose for 5 weeks before the endpoint evaluation of efficacy and side effects.
Outcomes	Psychosis: Brief Psychiatric Rating Scale (BPRS) psychosis subscale Number of responders for psychosis Death Discontinuation due to adverse events
Identification
Notes	Sponsorship source: Supported in part by grants MH‐44176, MH‐50038, and MH‐55735 from NIMH; grants AG‐07370, AG‐07232, and AG‐08702 from the National Institute on Aging; NIH grant RR‐00645; and the Charles S. Robertson Memorial Gift for Alzheimer’s Disease Research from the Banbury Fund
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information
Allocation concealment (selection bias)	Unclear risk	No information
Comparability of groups (selection bias)	Unclear risk	No information
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Raters and patients were blind to the study design throughout phase A and phase B.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	The blinded research psychiatrist (D.P.D.) who evaluated the patients was also in charge of treatment, including dose adjustment, at all time points in the study.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Drop‐out was 9% overall. Results of completers analyses were shown in detail. "Intent‐to‐treat analyses were also conducted for the phase A sample, carrying forward the last observation", but results were not shown in detail. Apparently, they were not very different.
Selective reporting (reporting bias)	Unclear risk	No protocol available
Other bias	High risk	In the initial 1‐week single‐blind phase, all patients received placebo. At the end of this week, the patients who still met the entry criteria were eligible for phase A