To the Editor.
—We read with interest the report by Dr Bowden and colleagues1 on the effectiveness of divalproex sodium compared with lithium and placebo for hospitalized, acutely manic patients. We, too, have observed the therapeutic benefits of this anticonvulsant for refractory bipolar and schizoaffective patients who failed previous trials with lithium and/or carbamazepine.2However, we would like to comment on methodological issues that may bear on the interpretation of their findings. First, the lithium group comprised a large group of prior treatment failures; 86% (n=31) of this group had received lithium in the past, and 48% reported ineffectiveness or intolerance. Thus, lithium may be serving as an active placebo rather than as a pharmacologic treatment for many in this group. Furthermore, the lithium group was only half as large as either the placebo group or the divalproex group. Although Bowden et al correctly observe that this smaller