Pertinent to the methods concerns of Dr. Soldani et al., we examined a large number of potentially confounding factors that might mediate or moderate the initial unadjusted results. We had to narrow their number by some procedure that would allow sufficient power in the regression to appropriately examine the main hypothesis. To increase the likelihood of finding potentially important factors, we considered all potentially important variables without correcting for multiple comparisons, compared the groups on these variables, and examined their relationship to relapse. Variables such as those raised by Dr. Soldani et al. (bipolar subtype, substance abuse history, or other treatments such as mood stabilizers and other relevant factors related to treatment group assignment) were not significantly different between the groups (χ2<0.80, df=1 or 3, p>0.70). In addition, there were no DSM-IV-defined rapid cyclers included in our analysis because none met the study entry criteria for sustained response.