News

       Zhong and Xu recently published a paper in the Journal of Anesthesia titled “Universal Target-Controlled Infusion of Propofol Based on the Marsh Model and Adjusted Input Weights” [1]. The aim of the study was to improve the performance of target-controlled infusion (TCI) in adjusting the gap between predicted propofol concentrations and actual plasma concentrations. They achieved this improvement by modifying the input weights in the Marsh model to mimic the TCI behavior of the Eleveld model [2], which is currently considered the best model. The Marsh model is used in the Diprifusor system [3, 4], the only TCI system available in Japan. To understand the concepts in Zhong and Xu’s study, it is necessary to understand how patient characteristics influence the pharmacokinetics (PK) of commonly used anesthetic drugs and how these effects, in turn, influence the predictive performance of PK and TCI modeling. This article uses propofol as an example to explain these concepts and provides additional information about remifentanil and fentanyl.
       Pharmacokinetic (PK) models based on the compartmental model concept incorporate a number of PK parameters, such as volume of distribution, clearance, and rate constant. Once a PK model has been constructed for a specific drug, changes in plasma concentrations of the drug and its effector substances can be simulated over time based on the drug’s dosing history. Furthermore, target controlled delivery (TCD) can be achieved using specialized algorithms and PK models for infusion pump control.
       Even for the same drug, its pharmacokinetics (PK, dose-concentration relationship) can vary depending on factors such as patient size and metabolic capacity. For example, administering 50 mg of propofol to patients weighing 50 kg and 100 kg will result in different plasma concentration profiles. To account for these patient-specific factors, covariates are included in the PK parameter values ​​for adjustment. For example, the Marsh model used in Diprifusor uses actual body weight as an input covariate. However, these adjustments have limitations. This is because, in principle, the predictive performance of PK models is best suited to patients with similar characteristics to the population from which these models were developed. Typically, PK models widely used in clinical practice are developed based on subjects with “standard” background factors. Therefore, in such cases, it is appropriate to use PK models developed based on “patient-specific” data. Furthermore, in recent years, representative pharmacokinetic models of anesthetics have been published that cover a wider range of patient characteristics, including obese patients, the elderly, and children [2, 5, 6]. When applied to infusion simulation or target control systems, these pharmacokinetic models can provide more accurate predictions for patients with characteristics similar to those used in model development.
       Currently, the only targeted controlled infusion (TCI) system approved for clinical use in Japan is the Diprifusor TCI system for propofol [7], which uses the Marsh model. The predictive performance of this system has been well validated in standard adults [3, 8]. The Marsh model only includes body weight as a covariate, and the dataset used to develop it did not account for extreme obesity [9]. Furthermore, gender and age were not included in the model parameters. Therefore, the plasma/action site drug concentration displayed on the pump may differ from the actual value (Figures 1 and 2).
       The solid line in the figure represents the propofol concentration at the action site as measured by the Diprifusor pump, and the dashed line represents the propofol concentration at the action site remodeled using the Eleveld model. The modeling data were obtained from a TCI dosing history. The patient was a 40-year-old man, 176 cm tall and weighing 150 kg. The target plasma propofol concentration during TCI was 3 μg/mL over 60 minutes. Modeling results based on the Eleveld model (including inclusion of severely obese individuals) were assumed to be more accurate than those based on the Marsh model. PCes is the propofol concentration at the action site.
       Based on target-controlled infusion (TCI) dosing data using the Diprifusor system, propofol concentrations at the action site were remodeled using the Eleveld model. The study included four patients aged 80 or 20 years, of either gender, with the same height and weight (170 cm, 68 kg). The target plasma propofol concentration for TCI was 3 μg/mL administered over 60 minutes. It was assumed that the modeling results based on the Eleveld model (including subjects with severe obesity) were closer to reality than those based on the Marsh model.
       In the above-mentioned limited conditions, when performing total intravenous anesthesia with emphasis on plasma or action site concentrations in special patients (e.g., morbidly obese or elderly patients), the following options may be considered:
       (1) Abandon target-controlled infusion (TCI) and manually adjust the dose based on pharmacokinetic (PK) modeling results: Regarding the choice of PK model, the Schneider model [10], used in many PK simulators, allows the input of gender, age, height, and weight. Furthermore, the Eleveld model [2] is capable of simulating a wider range of patients (from neonates to 88-year-old adults with a weight of up to 160 kg and a BMI of up to 50) and is considered one of the most robust PK models as of 2024.
       (2) Identification of qualitative differences that arise when using target-controlled Diprifusor infusion and the interpretation of these differences by anesthesiologists: Figure 1 shows that in severely obese patients, the action site concentration displayed on the pump may be underestimated, while the actual concentration may be higher. Furthermore, Figure 2 shows that differences based on gender and age lead to variations in the action site concentration. For the same target value for target-controlled infusion, the action site concentration is higher in older patients than in younger patients. In patients of the same age, the action site concentration is slightly lower in women than in men. Therefore, if the clinical effects observed during target-controlled infusion (e.g., clinical signs or EEG parameters) do not correspond to the action site concentration displayed on the pump, it should be understood that the prognosis may differ.
       (3) Improving the predictive performance by adjusting the covariates entered into the TCI system: Cortines et al. demonstrated that in obese patients, entering adjusted weight (note: different from the “adjusted weight” mentioned in Zhong and Xu’s Excel spreadsheet, see details below) into the Diprifusor TCI system can maintain clinically acceptable predictive performance.[11] Zhong and Xu[1] investigated how to enter virtual weight as a covariate into the Marsh pharmacokinetic model and adjust the pharmacokinetic parameters of plasma TCI by combining the adjusted bolus dose to mimic the behavior of TCI at the site of the Eleveld model, thereby achieving clinically acceptable predictive performance in a wide range of patients. They generated a large number of combinations of virtual weight and adjusted bolus dose by combining age, height, actual weight, and gender, and used an optimization algorithm to compare the simulation results with those of the Eleveld model. The Excel spreadsheet they provided uses a complex regression equation to calculate the virtual weight and adjusted bolus dose (Figure 3). Clinical use may require approval in accordance with institutional policies.
       (4) Real-time output of propofol dosing history from the Diprifusor TCI pump and resimulation using a more accurate pharmacokinetic model: In Japan, this can be done using the Schneider model using Dräger SmartPilot® View (Dräger Medical, Lübeck, Germany). Alternatively, although not possible in real time, exporting dosing data from the Diprifusor TCI pump and resimulating it using a standard pharmacokinetic simulator with a different pharmacokinetic model may technically provide comparable information.
       A screenshot of the Excel spreadsheet provided by Zhong and Xu [1]. A 40-year-old man weighing 150 kg and 176 cm tall is used as an example. To conduct a target-controlled infusion (TCI) mimicking the Eleveld model with a target effector concentration of 3 μg/mL, Marsh’s body weight (102 kg) calculated from the Excel spreadsheet was first entered into Diprifusor (a). Then, the CeT value (μg/mL) was set to 3 (b). In Diprifusor, the infusion-induced CpT value was set to 8.7 as the target plasma concentration (c). After the infusion started, when the plasma concentration predicted by Diprifusor reached the same value, the target plasma concentration was reset to 3. This adjustment brought the effector concentration closer to the value of 3 estimated by the Eleveld model. If the concentration at the action site needs to be increased by 1 μg/mL (e.g. from 3 μg/mL to 4 μg/mL), the target plasma concentration is increased in the TCI and after administration of 80 mg propofol at the dose specified in the Adjusted Push-In (d), the target plasma concentration is set at 4 μg/mL.
       Remifentanil is one of the anesthetic drugs whose dosage is usually adjusted based on body weight in clinical practice. For adult patients, the standard induction dose is 0.5–1 μg/kg/min, and the maintenance dose is 0.25 μg/kg/min. However, the Japanese prescribing information states that for obese patients with a BMI greater than 25, the dose should be determined based on ideal body weight rather than actual body weight. This suggestion may be related to the study by Egan et al. [12] on the pharmacokinetics of remifentanil in obese patients, which stated, “Because the pharmacokinetic parameters of remifentanil appear to be more closely related to lean body mass (LBM) than to total body weight (= actual body weight), the dose of remifentanil should be determined based on lean body mass (or ideal body weight) rather than total body weight.” Ideal weight and lean body mass are theoretical values ​​calculated using formulas and are often lower than the actual weight in obese patients [13]. These values ​​are used for dose adjustments, as the medication requirement in obese patients does not increase proportionally with body weight. It is worth noting that Egan’s study included patients with a BMI of up to 43, so this adjustment method may not be applicable to patients with more severe obesity.
       In fact, the standard Minto model for remifentanil[14] uses lean body mass (LBM) as a covariate, which may lead to inaccurate predictions for severely obese patients[5], since the LBM model was developed at a time when severely obese patients were relatively rare even in Western countries. This makes the LBM formula outdated. To address this issue, a method for adjusting the pharmacokinetic predictions of the Minto model for severely obese patients by introducing “virtual growth”[15] was proposed. This issue was subsequently resolved with the introduction of the Kim-Obara-Egan model[5]. This model covers obese patients with a BMI threshold of up to 73.7. Currently, this model is applied to various target-controlled infusion pumps marketed internationally.
       The pharmacokinetic model developed by Shafer et al. [16] is considered the gold standard for predicting plasma fentanyl concentrations. This model uses actual body weight as the only parameter-adjusted covariate. However, there are still some problems with the prediction accuracy in obese patients. Shibutani et al. [17, 18] proposed a method to improve the prediction accuracy of the Shafer model by incorporating pharmacokinetic weight (52/[1 + (196.4*e−0.025 TBW−53.66)/100]) into the model. This method was validated in obese patients weighing up to 181 kg.
       As demonstrated in this article, researchers have undertaken numerous efforts to improve the prediction accuracy of existing pharmacokinetic models and enhance their fit to actual data (i.e., to obtain more accurate pharmacokinetic model predictions). For example, they included virtual weight or height data as covariates in the model, which differ from actual patient data, making this particularly useful for patients whose characteristics fall outside the scope of existing models. Zhong and Xu’s study is a continuation of this line of research. Further rigorous validation across multiple institutions is needed to confirm the effectiveness of the model.
       Zhong G, Xu S. Universal target-controlled propofol infusion using the Marsh model with body weight adjustment. J Anesthesiology. 2024;38:275–8.
       Eleveld DJ, Colin P, Absalom AR, Struijs M. Pharmacokinetic-pharmacodynamic model of propofol and its widespread use in anesthesia and sedation. Br J Anaesth. 2018;120:942–59.
       Swinhoe KF, Peacock JE, Glen JB, Reilly KS. Predictive evaluation of the performance of the TCI Deprifusor system. Anesthesiology. 1998;53(Suppl 1):61–7.
       Marsh B, White M, Morton N, Kenny GN. A pharmacokinetic model guiding propofol infusion in children. British Journal of Anaesthesiology. 1991;67:41-8.
       Kim TK, Obara S, Egan TD, Minto KF, La Colla L, Drover DR, Wuijk J, Mertens M. Remifentanil disposition in obesity: a new pharmacokinetic model taking into account the influence of body weight. Anesthesiology. 2017;126:1019–32.
       Eleveld DJ, Proost JH, Vereecke H, Absalom AR, Olofsen E, Vuyk J, Struys M. Allometric growth model for remifentanil pharmacokinetics and pharmacodynamics. Anesthesiology. 2017;126:1005–18.
       Obara S, Kamata K, Nakao M, Yamaguchi S, Kiyama S. Practice guidelines for total intravenous anesthesia. Journal of Anesthesiology. 2024. https://doi.org/10.1007/s00540-024-03398-2.
       Ogawa T, Obara S, Akino M, Hanayama S, Ishido H, Murakawa M. Prognostic efficacy of targeted controlled infusion of propofol in robotic-assisted laparoscopic prostatectomy with carbon dioxide pneumoperitoneum in the head-down position. J Anesth. 2020;34:397–403.
       Gepts E, Camu F, Cockshott ID, Douglas EJ. Distribution of propofol in the human body following constant-rate intravenous administration. Anesthesia and Analgesia. 1987;66:1256–63.
       Schneider TV, Minto KF, Gambus PL, Andresen K, Goodale DB, Shafer SL, Youngs EJ. Effect of route of administration and concomitant factors on the pharmacokinetics of propofol in adult volunteers. Anesthesiology. 1998;88:1170–82.
       Cortines LI, De la Fuente N, Eleveld J, Oliveros A, Crovari F, Sepulveda P, Ibakache M, Solari S. Efficacy of a targeted controlled infusion model of propofol in obese patients: a pharmacokinetic and pharmacodynamic analysis. Anesthesiology Modeling. 2014;119:302–10.
       Egan TD, Huizinga B, Gupta SK, Jaarsma RL, Sperry RJ, Yee JB, Muir KT. Pharmacokinetics of remifentanil in obese and normal-weight patients. Anesthesiology. 1998;89:562–73.
       Obara S, Yoshida K, Inoue S. How obesity affects the administration of intravenous anesthetics. Curr Opin Anaesthesiol. 2023;36:414–21.
       Minto KF, Schneider TW, Egan TD, Youngs E, Lemmens HJ, Gambus PL, Billard W, Houk JF, Moore KH, Herman DJ, Muir CT, Mandema JW, Shafer SL. Effects of age and gender on the pharmacokinetics and pharmacodynamics of remifentanil: I. Model development. Anesthesiology. 1997;86:10–23.
       La Colla L, Albertin A, La Colla G, Porta A, Aldegheri G, Di Candia D, Gigli F. Predictive efficacy of the remifentanil Minto pharmacokinetic parameter panel in morbidly obese patients based on a new method for calculating muscle mass. Clinical Pharmacokinetics. 2010;49:131–9.
       Shafer SL, Varvel JR, Aziz N, Scott JK. Pharmacokinetics of fentanyl administered with a computer-controlled infusion pump. Anesthesiology. 1990;73:1091–102.
       Shibutani K, Inchiosa MA Jr, Sawada K, Bayramian M. Accuracy of pharmacokinetic models in predicting plasma fentanyl concentrations in normal-weight and overweight surgical patients: determination of administration mass (“pharmacokinetic mass”). Anesthesiology. 2004;101:603–13.
       Shibutani K, Inchiosa MA Jr, Sawada K, Bayramian M. Pharmacokinetic characteristics of fentanyl for postoperative analgesia in lean and obese patients. Br J Anaesth. 2005;95:377–83.
       The author would like to thank the Editorial Department of the English Department, Faculty of Science, Fukushima Medical University for editing this manuscript.
       Department of Anesthesiology, Pain Treatment and Surgery Center, Fukushima Medical University Hospital, Hikioka 1, Fukushima City, Fukushima Prefecture, Japan, 960-1295, Japan
       Springer Nature remains neutral with regard to territorial claims and institutional affiliations in published maps.
       Obara, S. The concept of “virtual body mass” in pharmacokinetic modeling and target-controlled infusion. Journal of Anesthesiology 38, 733–737 (2024). https://doi.org/10.1007/s00540-024-03413-6