Assessment of the performance of the Marsh model in effect site mode for target controlled infusion of propofol during the maintenance phase of general anaesthesia in an unselected population of neurosurgical patients.

BACKGROUND: Propofol target-controlled infusion (TCI) in effect site mode has become popular since it became commercially available. OBJECTIVE: We have performed a study to assess the pharmacokinetic performance of the Marsh model in effect site mode in an unselected group of patients during neurosurgery during the maintenance phase of anaesthesia. DESIGN: Fifty American Society of Anesthesiologists (ASA) physical status classes 1 to 3 adults underwent elective neurosurgery receiving propofol TCI using the Marsh model in effect site mode. Propofol dose titration and level of patient monitoring was determined by the attending anaesthesiologist. Arterial blood was sampled at regular intervals during the maintenance phase of anaesthesia and measured plasma propofol concentrations were compared with those estimated using TCI. SETTING: Large tertiary referral centre in Birmingham, UK, with a specialist neuroanaesthesia service. PATIENTS: Fifty ASA status I to III adult patients undergoing elective neurosurgery. MAIN OUTCOME MEASURES: Performance of Marsh model as assessed by median performance error (bias) and median absolute performance error (imprecision). RESULTS: Performance of the Marsh model showed a positive bias (median performance error) of 27.6%, and imprecision (median absolute performance error) of 29.4%. Analysis of pooled data demonstrated greatest bias in the early phase (15 to 30 min) of anaesthesia (mean prediction error of 51.6%). Analysis of covariates demonstrated that obesity (BMI >30 kg m(-2)) contributed around half of the bias detected (mean prediction error 47 vs. 23%, P < 0.001). Patients with advanced age and significant comorbidity (ASA physical status class >2) actually demonstrated significantly lower prediction errors. CONCLUSION: Pharmacokinetic analysis suggests that the performance of the Marsh model in effect site mode is poor in this broad patient population. The greatest bias demonstrated occurred in the early maintenance phase of anaesthesia. Of the covariates analysed, obesity contributed most significantly to an increased bias. Despite overall poor performance of the Marsh model, attending anaesthesiologists modified targeted propofol concentrations only 0.3 times per hour on average, using remifentanil dose modification nine times more frequently, with good surgical conditions in all patients.