In 2005, a meta-analysis conducted by the US Food and Drug Administration showed a 1% to 2% increase in mortality, in absolute values, in elderly dementia patients treated with so-called atypical neuroleptics than in patients not receiving neuroleptics. One placebo-controlled trial, two new meta-analyses, and several cohort studies of various sizes and designs have been published since 2005. The double-blind placebo-controlled trial showed a statistically significant decline in mortality when neuroleptic therapy (risperidone or haloperidol in most cases) was withdrawn. One of the meta-analyses showed excess mortality in patients receiving atypical neuroleptics compared with those not receiving neuroleptics. The other meta-analysis, focusing solely on risperidone, showed a higher risk of vascular death than in placebo-treated patients. Four very large cohort studies also showed a trend towards excess mortality with conventional neuroleptics. In practice, as all neuroleptics have negative risk-benefit balances in elderly dementia patients, it is best to avoid using them in these patients, if possible. If a neuroleptic is nonetheless prescribed, treatment should be for the shortest possible duration, at the minimum effective dose.