Mortality Risk in Patients With Schizophrenia Participating in Premarketing Atypical Antipsychotic Clinical Trials

Objectives: Given the concern that mortality rates may be increased in geriatric patients exposed to atypical antipsychotic agents, we assessed mortality rates for adult patients with schizophrenia assigned to an investigational antipsychotic (olanzapine, quetiapine, risperidone, or ziprasidone), a control antipsychotic (haloperidol or chlorpromazine), or placebo in preapproval clinical development programs to assess relative risk with atypical antipsychotics as compared to typical antipsychotics or placebo.

Method: We reviewed safety data (from clinical trials conducted from approximately 1982 to 2002) for 16,791 adult patients with schizophrenia (DSM-III or DSM-IV criteria) in U.S. Food and Drug Administration (FDA) Summary Basis of Approval (SBA) reports for 6 antipsychotic drugs. Mortality rates were calculated for each treatment group (investigational agent, active control, orplacebo) on the basis of patient exposure years (PEY) and gross mortality. We compared the differences in mortality rates between placebo and investigational agents, active controls, and all antipsychotic drugs combined using odds ratios.

Results: By PEY analysis, the mortality rate for patients assigned to placebo treatment was significantly higher (p < .05) than for either the investigational antipsychotic (OR = 0.23, 95% CI = 0.13 to 0.45) or the active control group (OR = 0.19, 95% CI = 0.08 to 0.45). Although rates based on gross mortality were also higher with placebo treatment, statistical significance was only seen when comparing patients assigned to placebo with those assigned to the active control antipsychotic group (OR = 0.35, 95% CI = 0.15 to 0.82).

Conclusions: Despite reported excess mortality with antipsychotic use in elderly patients with dementia, SBA data did not reveal a similar increased risk of antipsychotics in adult patients with schizophrenia. However, methodological limitations of the FDA SBA reports may affect the generalizability of these findings.