Are all commonly prescribed antipsychotics associated with greater mortality in elderly male veterans with dementia?

Objectives: To estimate mortality risk associated with individual commonly prescribed antipsychotics. Design: Five-year retrospective study. Setting: Veterans national healthcare data. Participants: Predominantly male, aged 65 and older, with a diagnosis of dementia and no other indication for an antipsychotic. Subjects who received an antipsychotic were compared with randomly selected controls who did not. Exposed and control cohorts were matched according to their date of dementia diagnosis and time elapsed from diagnosis to the start of antipsychotic therapy. Measurements: Mortality during incident antipsychotic use. Results: Cohorts who were exposed to haloperidol (n=2,217), olanzapine (n=3,384), quetiapine (n=4,277), or risperidone (n=8,249) had more comorbidities than their control cohorts. During the first 30 days, there was a significant increase in mortality in subgroups prescribed a daily dose of haloperidol greater than 1 mg (hazard ratio (HR)=3.2, 95% confidence interval (CI)=2.2-4.5, P<.001), olanzapine greater than 2.5 mg (HR=1.5, 95% CI=1.1-2.0, P=.01), or risperidone greater than 1 mg (HR=1.6, 95% CI=1.1-2.2, P=.01) adjusted for demographic characteristics, comorbidities, and medication history using Cox regression analyses. Greater mortality was not seen when a daily dose of quetiapine greater than 50 mg (HR=1.2, 95% CI=0.7-1.8, P=.50) was prescribed, and there was no greater mortality associated with a dose less than 50 mg (HR=0.7, 95% CI=0.5-1.0, P=.03). No antipsychotic was associated with greater mortality after the first 30 days. Conclusion: Commonly prescribed doses of haloperidol, olanzapine, and risperidone, but not quetiapine, were associated with a short-term increase in mortality. Further investigations are warranted to identify patient characteristics and antipsychotic dosage regimens that are not associated with a greater risk of mortality in elderly patients with dementia.