Objective: To determine the optimal strategy for converting stable bipolar patients from twice-daily divalproex delayed release (DR) to once-daily divalproex extended release (ER). Method: This prospective, open-label, crossover study compared 4 divalproex regimens in euthymic outpatients with bipolar I disorder (DSM-IV diagnosis confirmed by Mini-International Neuropsychiatric Interview). Serum valproic acid levels were collected 12, 16, 20, and 24 hours after the last bedtime dose of the following regimens: DR twice daily (DR b.i.d.) during week 1; total daily DR dose once daily (DR q.h.s.) during week 2; once-daily ER at equal daily DR dose (ER 1:1) during week 3; and once-daily ER with the dose increased by 500 mg (ER + 500) during week 4. Patients continued on ER + 500 for 4 additional weeks after the pharmacokinetic phase. Side effects and psychiatric symptoms were assessed at weeks 1 through 4, 6, and 8. Twenty-one patients were enrolled from July 2002 to July 2004. Results: Of the regimens tested, DR q.h.s. produced the widest fluctuations in valproic acid levels, with the highest 12-hour (82 μg/ml) and lowest 24-hour (44 μg/ml) levels. The ER + 500 dose was the only regimen that maintained the mean minimum valproic acid concentration above 50 μg/mL. Each regimen was well tolerated, and no significant changes in psychometric indices were observed. Conclusions: When converting stable bipolar patients from twice-daily divalproex DR to once-daily ER, we recommend increasing the total daily dose by 250 to 500 mg to ensure maintenance of therapeutic valproic acid levels.