Predictive performance of the modified Marsh and Schnider models for propofol in underweight patients undergoing general anaesthesia using target-controlled infusion

Background: In our preliminary study, the modified Marsh (M-Marsh) model caused an inadvertent underdosing of propofol in underweight patients. However, the predictive performance of the M-Marsh and Schnider models incorporated in commercially available target-controlled infusion (TCI) pumps was not evaluated in underweight patients. Methods: Thirty underweight patients undergoing elective surgery were randomly allocated to receive propofol via TCI using the M-Marsh or Schnider models. The target effect-site concentrations (Ces) of propofol were, in order, 2.5, 3, 4, 5, 6 and 2 μg ml^-1. Arterial blood samples were obtained at least 7 min after achieving each pseudo-steady-state. Results: A total of 172 plasma samples were used to determine the predictive performance of both models. The pooled median (95% confidence interval) biases and inaccuracies at a target Ce ≤ 3 μg ml^-1 were-22.6 (-28.8 to-12.6) and 31.9 (24.8-36.8) for the M-Marsh model and 9.0 (1.7-16.4) and 28.5 (21.7-32.8) for the Schnider model, respectively. These values at Ce ≥ 4 μg ml-1 were-9.6 (-16.0 to-6.0) and 24.7 (21.1-27.9) for the M-Marsh model and 19.8 (12.9-25.7) and 36.2 (31.4-39.7) for the Schnider model, respectively. Conclusions: The pooled biases and inaccuracies of both models were clinically acceptable. However, the M-Marsh and Schnider models consistently produced negatively and positively biased predictions, respectively, in underweight patients. In particular, the M-Marsh model showed greater inaccuracy at target Ce ≤ 3 μg ml^-1 and the Schnider model showed greater inaccuracy at target Ce ≥ 4 μg ml^-1. Therefore, it is necessary to develop a new pharmacokinetic model for propofol in underweight patients.