Predictive performance of a new pharmacokinetic model for propofol in underweight patients during target-controlled infusion

Background: In a previous study, the modified Marsh and Schnider models respectively showed negatively- and positively-biased predictions in underweight patients. To overcome this drawback, we developed a new pharmacokinetic propofol model-the Choi model-for use in underweight patients. In the present study, we evaluated the predictive performance of the Choi model. Methods: Twenty underweight patients undergoing elective surgery received propofol via TCI using the Choi model. The target effect-site concentrations (Ces) of propofol were 2.5, 3, 3.5, 4, 4.5, and 2 μg/mL. Arterial blood samples were obtained at least 10 minutes after achieving pseudo-steady-state. Predicted propofol concentrations with the modified Marsh, Schnider, and Eleveld pharmacokinetic models were obtained by simulation (Asan pump, version 2.1.3; Bionet Co. Ltd., Seoul, Korea). The predictive performance of each model was assessed by calculation of four parameters: inaccuracy, divergence, bias, and wobble. Results: A total of 119 plasma samples were used to determine the predictive performance of the Choi model. Our evaluation showed that the pooled median (95% CI) bias and inaccuracy were 4.0 (−4.2 to 12.2) and 23.9 (17.6-30.3), respectively. The pooled biases and inaccuracies of the modified Marsh, Schnider, and Eleveld models were clinically acceptable. However, the modified Marsh and Eleveld models consistently produced negatively biased predictions in underweight patients. In particular, the Schnider model showed greater inaccuracy at a target Ce ≥ 3 µg/mL. Conclusion: The new propofol pharmacokinetic model (the Choi model) developed for underweight patient showed adequate performance for clinical use.