Maintenance studies in bipolar disorder have received increased attention in recent years. The interest is driven by apparent contradictions between results of early placebo-controlled trials of lithium and recent open studies, as well as interest in a new group of drugs with mood-stabilizing properties. The multiple outcome indices that require attention in prophylactic bipolar disorder studies add a dimension not present in acute studies of bipolar disorder. We present the methodology of a recently completed randomized, double-blind, placebo-controlled, parallel-group comparison of divalproex and lithium. We examine the consequences of salient design features, along with their implications for future studies. A fundamental conclusion is that such maintenance studies should be designed and executed to emphasize enrollment of patients with relatively active, severe forms of the illness. This goal is not achieved simply, as inherent features of long-term, placebo-controlled studies drive recruitment and enrollment in the direction of patients with milder forms of bipolar disorder. Attention to the frequency of both manic and depressive episodes and the severity of an index manic episode may aid in the selection of patients most suitable for studies designed to achieve adequate statistical power.