Twenty-five RCTs and 4 meta-analyses were included in the review. It is not clear whether there was any overlap of studies in the meta-analyses, and thus whether any RCTs were included twice. The breakdown according to drug type was:
typical antipsychotics, 2 meta-analyses (covering 12 RCTs) and 2 further RCTs (1,237 participants);
atypical antipsychotics, 6 RCTs (2,261 participants);
antidepressants, 5 RCTs (429 participants);
mood stabilisers, 5 RCTs (342 participants);
cholinesterase inhibitors, 2 meta-analyses (covering 18 RCTs) and 6 further RCTs (8,764 participants);
memantine, 2 RCTs (656 participants).
Typical antipsychotics.
A meta-analysis of 7 RCTs concluded that antipsychotics (haloperidol, thioridazine, thiothixene, chlorpromazine, trifluoperazine, acetophenazine) were beneficial in 18% of patients and found no difference in efficacy between the drugs on neuropsychiatric symptoms. Another meta-analysis concluded that haloperidol (1.2 to 3.5 mg/day) reduced aggression but was associated with adverse effects. One study reported an effect of thioridazine on agitation, but the trial was of a poor quality. Another trial found perphenazine to have had no benefit.
Atypical antipsychotics.
Four of the 6 RCTs reported a significant beneficial effect. In these studies, doses of 5 to 10 mg/day olanzapine or 1.0 mg/day risperidone had modest effects on neuropsychiatric symptoms in patients with Alzheimer disease or vascular dementia. Several of the studies also reported a greater incidence of adverse events, including stroke, in groups receiving atypical antipsychotics.
Antidepressants.
A trial of citalopram found a small significant improvement in some neuropsychiatric symptoms. Trials of sertraline, fluoxetine and trazodone detected no significant effects on neuropsychiatric symptoms.
Mood stabilisers.
Three RCTs found that valproate was ineffective for treating neuropsychiatric symptoms and that it also caused adverse events. Carbamazide was beneficial in one trial and ineffective in another.
Cholinesterase inhibitors. One meta-analysis reported a statistically significant improvement in one neuropsychiatric symptom measure. However, this might have been due to the inclusion of trials on metrifonate, which had not been approved for use in the USA on account of its toxicity. Two studies reported a very small improvement with galantamine. Donepezil was found to be beneficial in 2 studies and ineffective in two others, including the longest trial of cholinesterase inhibitors. A trial of rivastigmine found no beneficial effect.
Memantine.
One trial found no significant effect on neuropsychiatric symptoms, while the other did not demonstrate an effect of clear clinical significance.