RESUMO
Target-controlled infusion systems are increasingly used to administer intravenous anaesthetic drugs to achieve a user-specified plasma or effect-site target concentration. While several studies have investigated the ability of the underlying pharmacokinetic-dynamic models to predict plasma concentrations, there are no data on their performance in predicting drug concentrations in the human brain. We assessed the predictive performance of the Marsh propofol model and Minto remifentanil model for plasma and brain tissue concentrations. Plasma samples were obtained during neurosurgery from 38 patients, and brain tissue samples from nine patients. Propofol and remifentanil concentrations were measured using gas chromatography mass spectrometry and liquid chromatography tandem mass spectrometry. Data were analysed from the nine patients in whom both plasma and brain samples were simultaneously obtained. For the Minto model (five patients), the median performance error was 72% for plasma and -14% for brain tissue concentration predictions. The model tended to underestimate plasma remifentanil concentrations, and to overestimate brain tissue remifentanil concentrations. For the Marsh model (five patients), the median prediction errors for plasma and brain tissue concentrations were 12% and 81%, respectively. However, when the data from all blood propofol assays (36 patients) were analysed, the median prediction error was 11%, with overprediction in 15 (42%) patients and underprediction in 21 (58%). These findings confirm earlier reports demonstrating inaccuracy for commonly used pharmacokinetic-dynamic models for plasma concentrations and extend these findings to the prediction of effect-site concentrations.