Fifty-nine patients with treatment-resistant depression were randomly allocated an addition of either lithium (Li; 31 patients) or carbamazepine (CBZ; 28 patients) to ongoing antidepressant treatment. The therapeutic efficacy of both strategies, assessed after 28 days, was not significantly different. In the Li group, clinical improvement significantly correlated with decrease in thyroxine concentration. Also, a decrease in cortisol level in Li responders and an increase in nonresponders was observed, suggesting a regulatory effect of Li on the hypothalamic-pituitary-adrenal axis activity. CBZ responders had lower baseline severity of depression than CBZ nonresponders and clinical improvement significantly correlated with increase in erythrocyte ATPase activity in the CBZ group.