Thoracic epidural analgesia (TEA) is effective in alleviating pain after major thoracoabdominal surgery and may also reduce postoperative mortality and morbidity. This study investigated cardiovascular autonomic control in patients undergoing elective thoracic surgery and its modulation by continuous TEA.Thirty-eight patients were randomly assigned to receive patient-controlled analgesia (PCA group) or thoracic epidural analgesia (TEA group) with doses of bupivacaine (0.25% during operation, 0.125% after operation) and fentanyl (2 microg ml(-1)). Heart rate variability (HRV), baroreflex function and pressure response to nitroglycerine and phenylephrine were assessed before operation, 4 h after the end of surgery (POD 0) and on the first and second postoperative days (POD 1 and POD 2).Early after surgery, all HRV variables and baroreflex sensitivities were markedly decreased in both groups. In the TEA group, total HRV and its high-frequency components (HF) increased towards preoperative values at POD 1 and POD 2, whereas the ratio of low to high frequencies (LF/HF) was significantly reduced (mean (SD), -44 (15)% at POD 0, -38 (17)% at POD 1, -37 (18%) at POD 2) and associated with blunting of the postoperative increase in heart rate and blood pressure. In the PCA group, the ratio of LF/HF remained unchanged and the decrements in HRV variables persisted until POD 2. In the two groups, baroreflex sensitivities and pressure responses recovered preoperative values at POD 2.In contrast with PCA management, TEA using low concentrations of bupivacaine and fentanyl blunted cardiac sympathetic neural drive, resulting in vagal predominance, while HRV variables were better restored after surgery.