BACKGROUND: This study evaluated the A-Line Autoregressive Index (AAI) response to surgical stimulation during lumbar arthrodesis, as an estimate of the anti-nociceptive component of a Bispectral Index (BIS) guided anaesthesia combined with epidural analgesia. METHODS: An epidural catheter was inserted in 23 patients allocated randomly to receive ropivacaine plus clonidine (Group R) or normal saline (Group S) epidurally. General anaesthesia was induced with propofol, cis-atracurium and a remifentanil infusion that was stopped 3 min after tracheal intubation, and maintained using sevoflurane to keep BIS at 50 (range 40-60). Mean arterial pressure, heart rate, end-tidal sevoflurane, BIS and AAI were analysed from 2 min before to 17 min after surgical incision. RESULTS: While BIS was maintained at 50, AAI significantly increased from a 2 min averaged value of 12 (4) to 21 (7) in Group S within the first 5 min after surgical incision, but did not change in Group R. Maximum AAI values reached during the study period were significantly higher in Group S than in Group R [38 (12) and 27 (10), respectively]. Binary logistic regression analysis allowed the calculation of AAI threshold values above which the probability of predominant nociception over anti-nociception was higher than 95%. At 1 MAC sevoflurane concentration, a 2 min averaged AAI of 35 or an AAI peak value of 62 were associated with such a probability. CONCLUSIONS: During a BIS-guided constant level of hypnosis, AAI response to the onset of surgical stimulation significantly differs according to the analgesic regimen. Further studies are needed to refine the estimation of sensitivity and specificity of this variable in assessing the balance between nociception and anti-nociception during general anaesthesia.