To the Editor: In reply to Drs. Kruszewski and Paczynski’s comments, let us consider each point.

• 1. The doses of mixed amphetamine salts were not ineffective. In fact, the study revealed efficacy for mixed amphetamine salts for the doses used compared to placebo. It is true that higher doses might have been even more effective.

• 2. We agree that higher doses of divalproex might have led to even greater benefits, although the doses and serum levels used were associated with a substantial rate of response of 80%.

• 3 and 4. We agree that the small group size and a study conducted at only one site, by definition, limited generalizability and also recommend replication studies. However, we demonstrated strong statistical significance with the group we used.

• 5. We agree that longer-term studies are needed to best evaluate long-term safety and outcome.

That 20% of the patients with bipolar disorder could not be stabilized while taking open-label divalproex is not particularly surprising. The response rate of 80% with open-label divalproex was substantial, however, and similar to what has been found in other open-label studies (1). The 14-day treatment with mixed amphetamine salts and placebo was long enough to establish clinical statistical significance. Most patients (23 of 29) did elect open treatment with mixed amphetamine salts for 6 additional weeks. We made no claims of efficacy for open divalproex treatment, only that it was associated with a benefit in this group. The elicited and spontaneously reported side effects in the entire trial were very low, perhaps because we did not aggressively "load" the divalproex and we used relatively low doses of mixed amphetamine salts in the crossover study. In the open extension (when the dosing of mixed amphetamine salts was not limited), the average dose remained low, at 14.5 mg/day, suggesting that this relatively low dose was clinically useful.

The use of the last observation carried forward is considered the most rigorous way to look at data from clinical trials. The divalproex responders were, in fact, all study completers. The only patient with a response who did not complete this phase of the study was one who improved so much during the first arm of the mixed amphetamine salts/placebo crossover study that the child’s mother did not want to risk a change in treatment. This patient was treated with mixed amphetamine salts outside the study and did very well.

As to adverse events, one other patient developed mania: this was clearly stated in the article and the abstract. There were no serious adverse reactions: this was clearly stated in the article. Three patients were hospitalized very early in the course of the open-label divalproex treatment, likely before these patients had adequate opportunity to respond to divalproex.

We do believe that this small, well-controlled study provides a basis for considering larger, more definitive and generalizable trials. Given the clear lack of efficacy of divalproex for ADHD symptoms and the positive effects of mixed amphetamine salts (versus placebo) in a randomized blinded comparison, we believe that such a combination approach (divalproex followed by mixed amphetamine salts) seems promising, and at least with the group and follow-up data that were available, reasonable tolerability and safety can be expected (at the doses used). We certainly believe that this first study should be followed by larger, more definitive controlled trials to better assess generalizability and tolerability in a larger group.