To the Editor: Regarding the study by Jan Volavka, M.D., Ph.D., et al. (1), the authors should be commended for attempting to differentiate four different antipsychotics in a single trial. However, I believe that the conclusions reached in this study cannot be supported by the data. During the last 6 weeks of the trial, the dose was increased in all four treatment groups. Only the olanzapine group had greater efficacy as a result. Why was the dose increased in the other treatment groups if there was no further improvement? In the case of risperidone, this increase resulted in a mean dose of 11.6 mg/day. It has been well established that risperidone doses above 10 mg/day are less effective than lower doses (2). Thus, this study demonstrated that investigators are unable to optimize patient response using dose titration. An alternative design (e.g., with a fixed dose) should have been employed.

The authors made little justification for the choice of olanzapine dose. The current labeling for olanzapine states that its antipsychotic efficacy occurs between 10 and 15 mg/day and that doses above 10 mg/day are not more efficacious. Despite this, the labeling was ignored, and the authors chose a target dose of 20 mg/day. Why did the authors design a trial in which patients were targeted with olanzapine at twice the recommended dose? Furthermore, why were the patients allowed to have their doses titrated up to 40 mg/day of olanzapine—more than twice the known safety limit? Fortunately, there were no serious adverse events during the trial. Further studies are needed to demonstrate that higher doses of olanzapine may be warranted and are safe. This should have been stated in the text.

In addition to problems with efficacy, there is also the issue of potential unblinding of the trial due to lack of tolerability. The authors attempted to mask the expected extrapyramidal symptoms of haloperidol by giving their patients prophylactic benztropine. A benztropine placebo was given for the other antipsychotics, and actual benztropine was blindly used only if needed. The labeling for risperidone indicates there is a dose-related increase in extrapyramidal symptoms. This becomes significantly higher than with placebo in doses of more than 10 mg/day. In the present trial, 32% of the patients taking risperidone required benztropine, compared to 13% for both the olanzapine and clozapine patients. For a rater observing extrapyramidal symptoms, the a priori likelihood that the patient was taking risperidone was significantly higher; as a result, it is conceivable that the blinding may have been compromised and the scoring biased.

In summary, by arbitrarily picking doses outside currently approved drug labeling, using a dose titration scheme that was unable to detect the maximally effective dose, and failing to adequately mask extrapyramidal symptoms, the authors designed a study that could not possibly have reached a conclusion as to which antipsychotic drug was superior to haloperidol. It is unfortunate that this trial belongs to the growing category of studies in which a flawed design yields uninterpretable results.