Septic shock that requires therapy with adrenergic agents is associated with high rates of mortality. Inappropriately normal or low serum concentrations of vasopressin contribute to the development of hypotension during sepsis. We critically evaluated the role of administering exogenous vasopressin to patients with septic shock. A computerized search of MEDLINE from January 1966--December 2003 and a manual search of relevant journals for abstracts were conducted. Eleven retrospective, six prospective cohort, and four prospective randomized studies were identified. Most studies evaluated short-term infusions of vasopressin at 0.08 U/minute or less as add-on therapy in patients requiring adrenergic agents. The results show that starting vasopressin in patients with septic shock increases systemic vascular resistance and arterial blood pressure, thus reducing the dosage requirements of adrenergic agents. These effects are rapid and sustained. Substantial enhancement of urine production, likely due to increased glomerular filtration rate, was shown in several studies. A few studies demonstrated clinically significant reduced cardiac output or cardiac index after vasopressin was begun, necessitating cautious use in patients with cardiac dysfunction. Vasopressin was associated with ischemia of the mesenteric mucosa, skin, and myocardium; elevated hepatic transaminase and bilirubin concentrations; hyponatremia; and thrombocytopenia. Limiting the dosage to 0.03 U/minut or less may minimize the development of these adverse effects. Vasopressin 0.03 U/minute or less should be considered if response to one or two adrenergic agents is inadequate or as a method to reduce the dosage of adrenergic agents. At present, vasopressin therapy should not be started as first-line therapy. Additional studies are needed to determine the optimum dosage, duration, and place in therapy of vasopressin relative to adrenergic agents. A multicenter, comparative study of vasopressin 0.03 U/minute as add-on therapy is under way and should provide mortality data.