To the Editor: As Dr. Nasrallah notes, our data provide confirmation of the higher mortality associated with haloperidol when compared with atypical antipsychotics in patients with dementia (1). The main finding and unique contribution of our paper, however, is that there are mortality risk differences between atypical antipsychotics, with risperidone and olanzapine having higher mortality rates than quetiapine. Since the publication of our article, our findings of differential mortality among individual antipsychotics have been confirmed in another sample (2).

Dr. Nasrallah also brings his pilot study (3) to our attention. This retrospective study at a single center reported higher rates of 2-year mortality for patients taking haloperidol in comparison with those taking atypical antipsychotics. However, as noted in a letter to the editor regarding that paper (4), the study did not control for the known selection biases that occur in patients treated with haloperidol compared with atypical antipsychotics. Haloperidol tends to be prescribed for patients older and sicker than those treated with atypical antipsychotics (5). In our study, we analyzed a wide array of potential confounding factors in addition to using propensity methods to control for potential treatment-by-indication bias. In doing so, we observed the mortality action of haloperidol occurring within the first 30 days of treatment. Therefore, it is unclear that one could conclude that neurotoxicity is the mechanism of mortality risk.

In light of our data, the evidence from randomized controlled trials, and a number of retrospective database studies, we find no support for the idea that atypical antipsychotics are neuroprotective in patients with dementia. Randomized trials have shown atypical antipsychotics to have 1%–2% higher risk than placebo over 10- to 12-week study periods (6). Over the longer 6-month follow-up in our cohort, olanzapine and risperidone showed mortality rates of approximately 27 deaths per 100 person-years of treatment compared with 18.6–21 deaths per 100 person-years with quetiapine and valproic acid. In addition, we previously showed (7) that the absolute mortality risk over 12 months in patients taking atypical antipsychotics was 4.8% higher than in those not taking medication, which corresponds to a number needed to harm of 20.8. Therefore, if atypical antipsychotics are to be prescribed, then they should be used in conjunction with a risk-benefit approach taking into account the efficacy and safety evidence base for the agents under consideration.