In the field of anesthesiology, propofol is a widely used intravenous anesthetic, and its pharmacokinetic (PK) and pharmacodynamic (PD) models are crucial for precise control of anesthetic depth. The Eleveld model is a widely applied PK-PD model for propofol, capable of predicting propofol plasma concentrations, effect-site concentrations, and the associated Bispectral Index (BIS) values. BIS is an index derived from processed electroencephalogram (EEG) signals to assess the depth of anesthesia and is widely used in clinical anesthesia.

Although the Eleveld model performs well in predicting propofol concentrations, its performance in predicting BIS values remains unclear. Particularly in Total Intravenous Anesthesia (TIVA), the accuracy of BIS predictions is critical for precise control of anesthetic depth. Therefore, this study aimed to evaluate the performance of the Eleveld model in predicting BIS values during TIVA and to investigate whether it can reliably predict actual measured BIS values.

Source of the Paper

This paper was co-authored by Ettienne Coetzee, Johan F. Coetzee (Jeff), and Marlis Haasbroek, affiliated with the Department of Anaesthesia and Perioperative Medicine at the University of Cape Town and the Department of Anaesthesiology and Critical Care at Stellenbosch University, respectively. The paper was published on August 23, 2024, in the British Journal of Anaesthesia (BJA), titled “Predictive pharmacodynamic performance of the Eleveld pharmacokinetic-pharmacodynamic model for propofol: comparison of predicted and measured bispectral index.”

Research Process

Study Subjects and Sample

The study included 40 healthy adults (ASA physical status 1-2) undergoing lower limb surgery, aged between 18 and 65 years. Exclusion criteria included body weight below 70% or above 130% of ideal body weight, neurological disorders, cardiac, renal, or hepatic dysfunction, myopathy or muscular dystrophy, potential airway problems, recent use of psychoactive medications, need for premedication, and impaired decision-making capacity.

Experimental Design and Procedure

1. Anesthetic Induction: All participants were unpremedicated. Before induction, researchers connected monitoring devices and established an 18G intravenous line. Monitoring included electrocardiography, pulse oximetry, non-invasive blood pressure, inspired oxygen partial pressure, capnography, BIS monitoring, and neuromuscular blockade monitoring (TOF monitoring).

2. Drug Infusion: Remifentanil was administered via Target-Controlled Infusion (TCI) with an initial effect-site concentration (Ce) of 3 ng/ml. Propofol was administered via manually controlled Target-Guided Infusion (TGI), with doses based on the Eleveld model and guided in real-time by two pharmacokinetic simulation software programs (PKPD Tools and StelSim).

3. BIS Value Prediction: The Eleveld model was used to predict BIS values, and the agreement between predicted and measured values was assessed using Bland-Altman analysis. Additionally, Bland-Altman analysis was performed on supplementary data provided by the Eleveld model.

Data Analysis

1. Prediction Error Calculation: Prediction Error (PE) and Absolute Prediction Error (APE) were calculated for each participant, along with the Median Prediction Error (MDPE) and Median Absolute Prediction Error (MDAPE).

2. Bland-Altman Analysis: Bland-Altman analysis was used to assess the agreement between predicted and measured values, calculating the mean bias and limits of agreement (LoA).

3. Supplementary Data Analysis: Bland-Altman analysis was performed on supplementary data from the Eleveld model to validate the reliability of the study results.

Key Results

1. Prediction Errors: Although the median prediction errors were small (MDPE of -1.9, MDAPE of 10), the error ranges were wide (MDPE range: -18.5 to 24.3, MDAPE range: 1.7 to 24.3). 47.8% of MDAPE values exceeded 10 BIS units, indicating significant prediction errors.

2. Bland-Altman Analysis: Bland-Altman analysis showed a small mean bias (-0.52 BIS units) but wide limits of agreement (-27.7 to 26.2). The limits of agreement for each participant did not meet the requirements for interchangeability, indicating substantial individual variability between predicted and measured values.

3. Supplementary Data Analysis: Bland-Altman analysis of the supplementary data from the Eleveld model yielded similar results, further validating the limitations of the Eleveld model in predicting BIS values.

Conclusion

Although the Eleveld model performs well in predicting propofol concentrations, its performance in predicting BIS values exhibits significant individual variability and uncertainty. The results suggest that BIS values predicted by the Eleveld model should be interpreted with caution, particularly when adjusting target-controlled infusion concentrations to achieve desired BIS values, as there is a substantial risk of error. Therefore, the Eleveld model should be used in conjunction with other monitoring methods in clinical practice to ensure precise control of anesthetic depth.

Research Highlights

1. Key Findings: The Eleveld model exhibits significant individual variability in predicting BIS values. Although the median prediction errors are small, the wide error ranges limit its clinical applicability.

2. Methodological Innovation: The study utilized two pharmacokinetic simulation software programs (PKPD Tools and StelSim) for real-time guidance, ensuring the accuracy and reliability of the experimental data.

3. Clinical Significance: The findings suggest that anesthesiologists should exercise caution when using the Eleveld model to predict BIS values, particularly in resource-limited settings where prediction errors may lead to inadequate control of anesthetic depth.

Research Value

This study provides important clinical validation of the Eleveld model’s performance in predicting BIS values, revealing its limitations in clinical applications. The results offer significant guidance for anesthesiologists in precisely controlling anesthetic depth during TIVA, especially in settings where BIS monitoring devices are unavailable. In such environments, the predicted BIS values from the Eleveld model should be interpreted in conjunction with other clinical indicators.