The predictive performance of the Marsh model for propofol target-controlled infusion during laparoscopic surgery...

This study aimed to investigate whether CO₂ pneumoperitoneum and liver retraction affect predictive performance of the Marsh pharmacokinetic model for propofol target-controlled infusion (TCI) during laparoscopic or robot-assisted gastrectomy.

Adult patients undergoing elective laparoscopic or robot-assisted gastrectomy were enrolled. Arterial blood samples were collected at predefined intraoperative time points, and plasma propofol concentration (Cp) was measured by high-performance liquid chromatography. Performance error (PE), median PE (MDPE), median absolute PE (MDAPE), wobble, and divergence were calculated. Associations between PE and procedural factors, time after anesthetic induction, and cardiac index (CI) were evaluated using linear mixed-effects models. Finally, an exploratory PE-correction equation was internally evaluated by leave-one-patient-out cross-validation.

A total of 155 measurements from 16 patients were analyzed. The overall MDPE and MDAPE were 10.2 and 20.4%, respectively. In the fully adjusted model, liver retraction was associated with a 10.0-percentage-point higher PE (95% confidence interval, 1.3 to 18.7), whereas each 1 L/min/m2 decrease in CI was associated with a 16.4-percentage-point increase in PE (95% confidence interval, 6.1 to 26.6). PE decreased by 3.3 percentage points/h after anesthetic induction. Pneumoperitoneum and head-up positioning were not independently associated with PE. The exploratory PE-correction equation improved prediction metrics in internal validation.

The Marsh model showed an overall positive bias within conventional MDPE and MDAPE thresholds. Higher PE was associated with liver retraction, lower CI, and earlier time after anesthetic induction, whereas pneumoperitoneum was not independently associated with PE after adjustment.