A novel advisory system, the Anesthesia Advisory Display (AAD) has been recently developed. It displays information about actual and future predictions of anesthetic drug concentrations mapped to clinical end-points such as BIS, basing on the actual infusion rate combined with population pharmacokinetic and interaction pharmacodynamic (PKPD) models. According to the large variability observed among patients in clinical end-points to the same dosing history, a population model could fail in predicting the real patient's behavior: an "individualized" model is then necessary. An on-line estimation of the model's parameter for the single patients has been developed, basing on the extended Kalman filter algorithm. The method has been tested on data of 40 patients from a previous clinical study and the prediction of the individual models have been compared with the population model's.