BACKGROUND: Data from 12-week placebo-controlled trials have led to mounting concerns about increased mortality in patients with Alzheimer's disease (AD) who are prescribed antipsychotics; however, there are no mortality data from long-term placebo-controlled trials. We aimed to assess whether continued treatment with antipsychotics in people with AD is associated with an increased risk of mortality. METHODS: Between October, 2001, and December, 2004, patients with AD who resided in care facilities in the UK were enrolled into a randomised, placebo-controlled, parallel, two-group treatment discontinuation trial. Participants were randomly assigned to continue with their antipsychotic treatment (thioridazine, chlorpromazine, haloperidol, trifluoperazine, or risperidone) for 12 months or to switch their medication to an oral placebo. The primary outcome was mortality at 12 months. An additional follow-up telephone assessment was done to establish whether each participant was still alive 24 months after the enrollment of the last participant (range 24-54 months). Causes of death were obtained from death certificates. Analysis was by intention to treat (ITT) and modified intention to treat (mITT). This trial is registered with the Cochrane Central Registry of Controlled Trials/National Research Register, number ISRCTN33368770. FINDINGS: 165 patients were randomised (83 to continue antipsychotic treatment and 82 to placebo), of whom 128 (78%) started treatment (64 continued with their treatment and 64 received placebo). There was a reduction in survival in the patients who continued to receive antipsychotics compared with those who received placebo. Cumulative probability of survival during the 12 months was 70% (95% CI 58-80%) in the continue treatment group versus 77% (64-85%) in the placebo group for the mITT population. Kaplan-Meier estimates of mortality for the whole study period showed a significantly increased risk of mortality for patients who were allocated to continue antipsychotic treatment compared with those allocated to placebo (mITT log rank p=0.03; ITT p=0.02). The hazard ratio for the mITT group was 0.58 (95% CI 0.35 to 0.95) and 0.58 (0.36 to 0.92) for the ITT population. The more pronounced differences between groups during periods of follow up longer than 12 months were evident at specific timepoints (24-month survival 46%vs 71%; 36-month survival 30%vs 59%). INTERPRETATION: There is an increased long-term risk of mortality in patients with AD who are prescribed antipsychotic medication; these results further highlight the need to seek less harmful alternatives for the long-term treatment of neuropsychiatric symptoms in these patients. FUNDING: UK Alzheimer's Research Trust.