Dr. Jeste and Colleagues Reply

To the Editor: We thank Dr. Chaplin for his thoughtful comments on our article. We reported the results of an open-label study in which 330 institutionalized patients with moderate to severe dementia (mean age=82.5 years), complicated by psychosis or severe agitation, were treated with risperidone for up to 1 year. With a mean dose of 1 mg/day, the 1-year cumulative incidence of persistent tardive dyskinesia, calculated by use of Kaplan-Meier survival analysis (1), was 2.6%—considerably lower than that expected with conventional neuroleptics in an elderly group with dementia.

Dr. Chaplin raises questions about the dropout rate and the use of statistical analysis. Of the 330 patients who entered the study, 133 (40.3%) completed the 1-year investigation, whereas 197 (59.7%) discontinued treatment prematurely. The reasons for discontinuation (with percentages of the total group) were as follows: adverse events (N=87, 26.4%), voluntary discontinuation or administrative reasons (N=77, 23.3%), inadequate response (N=15, 4.5%), or intercurrent illness or abnormal laboratory results (N=18, 5.5%). The types of adverse events ranged from rigidity, agitation, and depression to cellulitis, pneumonia, and, in 8.8% of the patients (N=29), death. Given the advanced age of the subjects and the severity of their cognitive impairment, it is not surprising that a substantial proportion of these institutionalized patients experienced adverse events, whether related to the study medication or not. Because of the design of the investigation, our report did not provide good evidence to address causal associations of any of these events with risperidone therapy.

Survival analysis is a standard statistical technique employed in prospective longitudinal studies for the cumulative incidence of conditions such as tardive dyskinesia (2–4; Woerner et al., 1998). In long-term follow-up investigations, subjects may drop out at different time points. The Kaplan-Meier method (1) is one of the most commonly used techniques that allows use of “at least this long” information on all of the subjects who entered the study to prepare unbiased survival curve estimates. Survival analysis takes the “survival” times for a group of subjects and generates a survival curve, which shows the proportion of members who survive or remain “alive” over time (5). In this study, “survival” referred to the likelihood of a patient’s surviving without developing tardive dyskinesia.