Comparison of propofol pharmacokinetic and pharmacodynamic models for awake craniotomy: A prospective observational study

Anaesthesia for awake craniotomy aims for an unconscious patient at the beginning and end of surgery but a rapidly awakening and responsive patient during the awake period. Therefore, an accurate pharmacokinetic/pharmacodynamic (PK/PD) model for propofol is required to tailor depth of anaesthesia. OBJECTIVE To compare the predictive performances of the Marsh and the Schnider PK/PD models during awake craniotomy. DESIGN A prospective observational study. SETTING Single university hospital from February 2009 to May 2010. PATIENTS Twelve patients undergoing elective awake craniotomy for resection of brain tumour or epileptogenic areas. INTERVENTION Arterial blood samples were drawn at intervals and the propofol plasma concentration was determined. MAIN OUTCOME MEASURES The prediction error, bias [median prediction error (MDPE)] and inaccuracy [median absolute prediction error (MDAPE)] of the Marsh and the Schnider models were calculated. The secondary endpoint was the prediction probability P_k, by which changes in the propofol effect-site concentration (as derived from simultaneous PK/PD modelling) predicted changes in anaesthetic depth (measured by the bispectral index). RESULTS The Marsh model was associated with a significantly (P=0.05) higher inaccuracy (MDAPE 28.9±12.0%) than the Schnider model (MDAPE 21.5±7.7%) and tended to reach a higher bias (MDPE Marsh-11.7±14.3%, MDPE Schnider-5.4±20.7%, P=0.09). MDAPE was outside of accepted limits in six (Marsh model) and two (Schnider model) of 12 patients. The prediction probability was comparable between the Marsh (P_K 0.798±0.056) and the Schnider model (P_K 0.787±0.055), but after adjusting the models to each individual patient, the Schnider model achieved significantly higher prediction probabilities (P_K 0.807±0.056, P=0.05). CONCLUSION When using the 'asleep-awake-asleep' anaesthetic technique during awake craniotomy, we advocate using the PK/PD model proposed by Schnider. Due to considerable interindividual variation, additional monitoring of anaesthetic depth is recommended.