Twenty-nine RCTs were included in the review (4,954 participants, range 14 to 625). High risks of bias were found for funding sources of bias (14 trials), selective outcome reporting (one trial) and incomplete outcome data (one trial). Risks of bias for all other domains were rated as low or unclear (fully reported in paper).
Change in neuropsychiatric symptoms (29 trials): Compared with placebo, statistically significant improvements in neuropsychiatric symptoms were reported by some trials that assessed the efficacy of risperidone (0.5mg/day to 2mg/day; two of six trials) and olanzapine (5mg/day to 10mg/day; two of four trials). Similar results were reported by singular trials of aripiprazole (significance result not reported), carbamazepine (300mg/day), oestrogen (0.625mg to 2.5mg), propanolol (30mg/day to 120mg/day) and prazosin (1mg/day to 6mg/day) (each compared with placebo) and cyproterone (100mg/day) (compared with haloperidol).
Withdrawals, adverse events and mortality: Twenty-four trials assessed rates of all-cause withdrawals, 21 assessed withdrawals due to adverse events and 23 trials assessed mortality. Comparisons with placebos revealed statistically significantly higher rates of all-cause withdrawals for singular trials of 2mg/day risperidone, 2mg/day to 15mg/day aripiprazole, 6mg/day to 12mg/day rivastigmine and 50mg to 100mg quetiapine. A similar, statistically non-significant result was shown when oxcarbazepine (300mg/day to 900mg/day) and placebo were compared. The rate of all-cause withdrawal was statistically significantly lower with propranolol (30mg/day to 120mg/day) than with placebo. Statistically significantly higher rates of withdrawals due to adverse events were shown by singular trials for 2mg/day risperidone, 15mg olanzapine and 10mg/day aripiprazole than with placebo. One study reported that mortality was significantly higher with 1mg/day risperidone than with placebo.