Sixty ASA Grade II-III patients, without clinical symptoms of ventricular dysfunction and scheduled for elective bifemoral grafting for abdominal aneurysms, were allocated randomly to three equal groups to receive 150 micrograms sufentanil intravenously, epidurally or intrathecally. All patients received midazolam, vecuronium and nitrous oxide. Except for right ventricular stroke work index and pulmonary vascular resistance, all haemodynamic measurements decreased after sufentanil injection, but to the greatest extent after intravenous injection. Response to abdominal incision reversed the haemodynamic changes, although this was not accompanied by an increase in heart rate and coronary perfusion pressure after intrathecal injection. Four patients given intravenous, and four patients given epidural sufentanil required additional sufentanil after abdominal incision. Despite the changes in systemic vascular resistance and the concordant alterations in cardiac index after aortic cross-clamping and revascularization, intrathecal sufentanil provided more stability in heart rate than intravenous or epidural sufentanil. In conclusion, intrathecal sufentanil produced the most stable heart rate. Two patients in the intrathecal group developed spinal headaches post-operatively.