Hollis_2007_Am.J.Geriatr.Psychiatry_15_932

OBJECTIVE: To establish the instantaneous relative risk (RR) of death associated with individual antipsychotic drugs, carbamazepine and sodium valproate for those 65 years and older.
METHODS: Subjects dispensed antipsychotic drugs, sodium valproate or carbamazepine in 2003 or 2004 were analyzed as incident (N = 16,634) or prevalent (N = 9,831) users. Survival curves, mortality rates, and Cox proportional hazards models over two time periods were used to explore risk of death. The models were adjusted for age, sex, residential status, and psychotropic and medical drug dispensing. Olanzapine subjects were the reference group in the Cox regression. Subanalyses were performed for incident subjects with more than 30 days of follow-up and those dispensed cholinesterase inhibitors.
RESULTS: In the adjusted Cox proportional hazards models, haloperidol dispensing was consistently associated with an increased risk of death compared with olanzapine users (relative risk [RR] for incident users: 2.26, 95% confidence intervals (CI): 2.08-2.47; Wald statistic: 345.36, df = 1, p < or =0.001). There was some evidence of decreased survival with dispensing of higher haloperidol doses, although confounding by medical comorbidity cannot be excluded. Chlorpromazine (RR: 1.39, 95% CI: 1.15-1.67; Wald statistic: 12.08, df = 1, p <0.001) and risperidone (RR: 1.23, 95% CI: 1.07-1.40; Wald statistic: 9.12, df = 1, p = 0.003) dispensing were associated with increased risk of death in incident users. CONCLUSION: These results should be interpreted cautiously because haloperidol and chlorpromazine are used in broader clinical contexts. However, in the absence of data from randomized trials, the safety profile of haloperidol should not be assumed to be benign. Antipsychotic drugs should not be studied as an aggregated group because their associated risks are not uniform.