To the Editor:

Dr. Adetunji and colleagues object to reporting secondary findings of clinical trials, implying that this could be misleading and that such data should be part of the primary publication. We made clear in both abstract and text that these were secondary analyses of a previously published clinical trial. Multisite, large-scale trials often take several years to carry out and produce a wealth of data; only a small part can be included in the primary report. Indeed, major journals impose strict word limits that allow only the main outcome variables to be published. Moreover, including secondary outcomes in a primary report could distract from the primary hypothesis. On the other hand, ignoring secondary outcome variables would mean missing potentially important information.

The power analysis was based on the ability to detect a clinically significant difference between risperidone and placebo groups in the primary outcome. Nevertheless, we are interested in other outcomes as well. Full details on methodological or design issues have been published (1). The decision to use a parent-rated version of the Ritvo-Freeman Real Life Rating Scale was dictated by an inability to reach acceptable interrater reliability among clinician observers despite intensive training. We do not claim that the parent-rated version is more valid, but it may be a better way of scoring real-life functioning in that the parents sampled a larger, more representative portion of child behavior.

Dr. Adetunji et al. suggest that a reduction in irritability and aggression could result from drug-induced fatigue and drowsiness, a possibility worth discussing because of the mistaken belief in some quarters that antipsychotic drugs cause cognitive impairment. These adverse effects, although prevalent across patients, were nevertheless transient for each patient, while improvement was sustained well beyond these side effects (2). In no case was sedation considered an acceptable mediator of benefit: if sedation continued for more than a week, dose adjustments were made that usually resolved it.

Dr. Adetunji and colleagues object to our mentioning a "nearly significant difference" of p<0.07. Although this value does not meet the conventional cutoff for statistical significance, it may be imprudent simply to dismiss it.

In conclusion, the data from this placebo-controlled trial indicate that risperidone improved the restricted, repetitive, and stereotyped behaviors that constitute one of the three core symptom clusters of autism in addition to decreasing aggression, self-injury, and severe tantrums. These data do not undermine the value of behavior therapy in the treatment of autism. We are currently conducting a clinical trial to study the effects of combining medication and behavior therapy in the treatment of pervasive developmental disorders accompanied by serious behavioral problems.