Open Access
27.10.2024 | Original Research Article
Population Pharmacokinetic Modelling of Norepinephrine in Healthy Volunteers Prior to and During General Anesthesia
verfasst von:
Yingxue Li, Jeroen V. Koomen, Douglas J. Eleveld, Johannes P. van den Berg, Jaap Jan Vos, Ilonka N. de Keijzer, Michel M. R. F. Struys, Pieter J. Colin
Intraoperation hypotension (IOH) is commonly observed in patients undergoing surgery under general anesthesia, and even a brief episode of IOH can lead to unfavorable outcomes. To reduce the risk, blood pressure is closely measured during general anesthesia, and norepinephrine (NE) is frequently administered if hypotension is detected. Despite its routine application, information on the dose-exposure-response relationship of NE remains limited. Additionally, quantification of the influence of general anesthesia on the pharmacokinetics (PK) of NE is lacking.
Objective
In this study, we aimed to describe NE PK in healthy volunteers and the influence of general anesthesia on its PK.
Methods
A single-center, cross-over study was conducted in healthy volunteers. The volunteers received a step-up NE dosing scheme (0.04, 0.08, 0.12, 0.16 and 0.20 mcg–1/kg–1/min–1) first in the awake state and then under general anesthesia. General anesthesia was administered using a propofol/remifentanil Eleveld target-controlled infusion. During general anesthesia, a 30-second electrical stimulus was given as surrogate for surgical incision to the volunteers at each dosage step. Blood samples were drawn before the initial dosing and after each dosing step, and plasma NE, propofol and remifentanil concentrations were subsequently determined. A population PK model was developed using non-linear mixed effects modelling. Simulations were conducted to predict the plasma NE concentration in patients at different measured propofol concentrations.
Results
A total of 1219 samples were analyzed from 36 volunteers. A two-compartment model with a first-order elimination best described the data. Weight, age, and session effect (awake vs general anesthesia) were identified as relevant covariates on the clearance (CL) of NE. A 10% decrease in NE CL was observed after general anesthesia induction. This difference between sessions is better explained by the measured concentration of propofol, rather than the anticipated impact of cardiac output. The estimated post-stimulation NE concentration is 0.66 nmol/L–1 (95% CI 0.06–1.20 nmol/L–1) lower than the pre-stimulation NE concentration. Model simulation indicates that patients at a higher measured propofol concentration (e.g., 6 mcg/mL–1) exhibited higher NE concentrations (95% PI 18.10–43.89 nmol/L–1) than patients at a lower measured propofol concentration (e.g., 3 mcg/mL–1) (95% PI 16.81–38.91 nmol L–1).
Conclusion
The NE PK is well described with a two-compartment model with a first-order elimination. NE CL exhibiting a 10% decrease under general anesthesia, with this difference being attributed to the measured concentration of propofol. The impact of stimulation on NE PK under general anesthesia is very limited.
Norepinephrine pharmacokinetics in healthy volunteers is characterized by a two-compartment model with first order elimination.
Norepinephrine clearance is 10% lower during general anesthesia than during the awake phase. This difference is better explained by the measured concentration of propofol rather than by cardiac output. Individual variability exists in the effect of measured concentration of propofol on norepinephrine clearance.
After intermittent noxious tetanic stimulation, a decrease in norepinephrine concentration was observed. However, based on the estimated magnitude of this reduction, it is unlikely to cause clinically relevant effects.
1 Introduction
Intraoperation hypotension (IOH) is frequently encountered among patients undergoing surgery under general anesthesia [1, 2]. Intraoperation hypotension has various definitions; however, the occurrence of adverse outcome increases when mean arterial pressure (MAP) is below 60–70 mmHg even for a short period of time [3, 4]. The postoperative complications associated with IOH include myocardial injury, acute kidney injury, and even death [4]. Intraoperation hypotension can be caused by various factors; however, the hypotension induced by the usage of anesthetics account for 30%–67% of IOH among all factors [5, 6]. Therefore a better management of blood pressure under general anesthesia is crucial in improving patient outcomes.
Anzeige
Norepinephrine (NE) is routinely used during general anesthesia as a vasopressor to treat arterial hypotension as it can quickly restore vascular tone and blood pressure [7, 8]. Despite its routine use, little is known on the dose-exposure-response relationship of NE and initial dosing largely depends on clinical experience. The pharmacokinetic (PK) properties of NE have been studied in adult patients with septic shock or traumatic brain injury, and in critically ill children [9, 10]. Previously, NE PK has been described with a one-compartment model, with variability in PK being accounted for by factors such as severity of illness and body weight. However, current knowledge regarding NE PK in the general adult population of patients presenting for elective surgery under general anesthesia is lacking.
The induction of general anesthesia is often accompanied by hemodynamic changes such as a decrease in cardiac output (CO) [11]. Given that NE is an intermediate-to-high extraction ratio drug, its CL is expected to be influenced by blood flow [12, 13]. Therefore, an interaction between NE PK and general anesthesia would be anticipated. Nevertheless, the exact nature of this impact still remains unclear. Moreover, surgical incision may also affect NE PK, as it typically triggers strong sympathoadrenal responses and the release of NE throughout the body when people are awake, often resulting in an increase of MAP and heart rate (HR) [11]. However, the impact of surgical incision on NE PK remains unclear in the presence of anesthetics, as high-dose opioids such as remifentanil suppresses sympathetic responses [14].
In this study, we aimed to describe NE PK in healthy volunteers and the influence of general anesthesia on its PK.
2 Material & Method
2.1 Study Design
A single-center cross-over study in healthy volunteers was conducted at the University Medical Center Groningen. The study was approved by the Medical Ethics Committee (Stichting Beoordeling Ethiek Biomedisch Onderzoek, METc Assen: July 21st, 2021, NL76998.056.21) and was registered in the Netherlands Trial Register (https://trialsearch.who.int, reference: NL9312). All volunteers provided written informed consent before the study procedures. The main results are described by our colleagues de Keijzer et al [15, 16]. The important aspects of the clinical trial are summarized below.
Anzeige
2.1.1 Inclusion and Exclusion Criteria
Thirty-six healthy volunteers were included in three groups according to age (18–34, 35–50 and 51–70 years) of equal proportion and included a 1:1 male-to-female ratio. Volunteers with the following conditions were excluded: pregnancy, cardiovascular disease, a difference of >15 mmHg between the systolic blood pressure (SBP) or diastolic blood pressure (DBP) measured on the left and right upper arm, increased risk of difficult mask ventilation or tracheal intubation, pulmonary, gastric or endocrinologic disease, end-stage kidney or liver failure, use of tricyclic antidepressants or monoamine oxidase (MAO) inhibitors.
2.1.2 Study Procedures
Standard monitoring was applied to all volunteers including electrocardiography (ECG), pulse-oximetry, non-invasive blood pressure and bispectral index (BIS) (Philips IntelliVue MP50 [Philips, Eindhoven, The Netherlands]; BIS VISTA, Aspect Medical Systems Inc., Norwood, USA). For each volunteer, a peripheral intravenous catheter was placed and a continuous infusion of Ringers Lactate (Baxter B.V., Utrecht, The Netherlands) was initiated at a rate of 10 mL/h–1. Under local anesthesia, a 20G arterial catheter was inserted in the radial artery for every volunteer. A computer system running RUGLOOP II software (Demed, Sinaai, Belgium) was used to synchronize data recording.
The study consisted of two phases: an awake phase and a general anesthesia phase. In the awake phase only NE was administered, whereas in the general anesthesia phase NE was administered in the concomitant presence of propofol and remifentanil. A washout period of 30 minutes was implemented. The general anesthesia phase commenced either when arterial blood pressure (ABP) returned to baseline level (within a 10% variability) or after one hour after the washout phase, whichever occurred first. Preoxygenation was performed for every volunteer using an inspired oxygen fraction of 1.0. Throughout the general anesthesia phase, each volunteer was fitted with a laryngeal mask, and mechanical ventilation was initiated using pressure-controlled ventilation. Tidal volumes of 6–8 mL/kg–1 (ideal body weight) were maintained, and zero positive end-expiratory pressure was applied using the Zeus ventilator (Dräger, Lübeck, Germany) to ensure normocapnia. Surgical incision was simulated by applying a nociceptive electrical stimulus during the general anesthesia phase. The stimulus was generated by EZstim III peripheral nerve stimulator (Halyard, Zaventem, Belgium), lasting 30 seconds (50mA, 100Hz) was administered during the general anesthesia phase following the sampling procedure at each incremental step. After the stimulus, another NE sample was drawn to assess the subclinical response it has triggered (i.e., the release of NE).
2.1.3 Drug Administration
During the awake phase, each volunteer first received a small bolus administration of 5 mcg then a step-up dosing regimen of NE (norepinephrine tartrate CF, Centrafarm, Breda, The Netherlands). The dosing regimen comprised five consecutive steps with doses of 0.04, 0.08, 0.12, 0.16, and 0.20 mcg/kg–1/min–1, each step lasting for 15 minutes. Norepinephrine was administered through peripheral intravenous infusion, utilizing a dilute solution of 20 mcg/mL–1 delivered via a BD Alaris GH plus syringe pump (Becton Dickinson, Eysins, Switzerland). During the procedure, an increase in SBP exceeding 200 mmHg for more than one minute was avoided at all times and was the per protocol reason for ceasing administration of norepinephrine. In addition, if mean arterial pressure (MAP) increased by more than 40% from baseline, the current step was finished and the NE dose was not further increased.
In the general anesthesia phase, propofol (Propofol MCT/LCT 20 mg/mL–1, Fresenius Kabi, Bad Homburg, Germany) and remifentanil (Remifentanil 50 mcg/mL–1 Mylan, Canonsburg, USA) were administered using target controlled infusion (TCI). Propofol dosing was guided by the Eleveld model to attain a 50% age-adjusted drug effect concentration [17] and the Eleveld remifentanil model was used to target an effect-site concentration of 3.6 ng/mL–1 [18]. The dosing combination was chosen to simulate anesthetic doses that are comparable to clinical practice and were shown to be favorable to obtain adequate anesthesia in terms of hemodynamics, noxious stimulation (i.e., movement and electroencephalogram suppression) [19]. Drug administration was managed using BD Alaris GH pumps (Cardinal Health, Basingstoke, UK) controlled by RUGLOOP II software (DEMED, Sinaai, Belgium). Fifteen minutes after general anesthesia induction, a NE bolus of 5 mcg was planned to be administered, followed by the initiation of a step-up NE dosing regimen that is the same as that used in the awake phase. In case MAP dropped below 50 mmHg after general anesthesia induction, NE infusion was initiated immediately instead of waiting for the intended starting point. Following the completion of the final dosing step, the administration of propofol and remifentanil was terminated. Administration of NE was gradually reduced with the goal of maintaining normotension.
2.1.4 Bioanalysis and Sample Collection
In the awake phase, a baseline blood sample was obtained from each volunteer prior to the initiation of any drug administration. A blood sample was collected at the end of every step of the step-up NE dosing scheme. After NE infusion was stopped, more samples were taken at 2, 5, and 30 minutes. During the general anesthesia phase, sample collection was conducted in a similar manner as in the awake state. Additional samples were collected upon cessation of anesthesia and subsequently at 2, 5, 10, 20, 30 minutes thereafter. Following the cessation of the stimulus, an additional blood sample was obtained from each volunteer. Following blood sample collection, the samples were processed and analysed according to previously described methods [20].
2.2 Population PK Modelling
Norepinephrine concentrations were analyzed by developing a non-linear mixed-effect population pharmacokinetic model. Data were fitted using the first-order conditional estimation method with interaction (FOCE-I) algorithm in NONMEM (version 7.5; Icon Development Solutions, Hannover, MD, USA) in combination with Pirana (version 3.0.0; Certara, Lugano, Ticino, Switzerland) as a graphical interface. The model building process started by exploring one-, two- and three-compartmental PK models with first-order elimination. As NE is an endogenous substance, a baseline level of NE, prior to NE administration, was described in the model according to Eq. 1.
where baseline denotes the amount of NE in the central compartment before NE administration. Prod (ng/min–1), CL (L/min–1) and Vc (L) denote the endogenous production rate, the clearance and the volume of distribution of the central compartment of NE. To account for the short delay between the time NE travels from the drug administration site to the sampling site in the contralateral arm, we explored the possibility of employing a lag time in the model.
Inter-individual variability (IIV) on the typical population parameters was assumed to be log-normally distributed using an exponential transform of the random effect. Bodyweight (BW) was included in the model a priori using allometric scaling [21]. In accordance with theory based allometric scaling, the exponents were 0.75 for clearance parameters and production rate, 1.00 for volumes of distribution parameters, and 0.25 for the lag time parameter. Age, weight, height, sex, body mass index (BMI), session effect (whether individuals were awake or under general anesthesia) were evaluated by analyzing the correlations between estimated IIV of the population parameters after the basic model was established. Covariates with a correlation coefficient >0.5 were tested in the model using linear, power or exponential functions. Relevant hemodynamic (CO, HR, MAP) and anesthetic (measured propofol concentrations, remifentanil concentrations) variables were explored to account for the session effect should one exist. Proportional, additive and combined error models were explored to describe residual variability.
2.3 Model Evaluation
For each step of model building, a modification of the model is accepted when: (1) the objective function value (as implemented in NONMEM) showed a reduction greater than 3.85, which corresponds to the 5% significance level critical quantile of the chi-square distribution; (2) improvements were seen in the goodness-of-fit plots; (3) precision and plausibility of the estimated parameters were considered reasonable.
Goodness-of-fit of the model was graphically evaluated in R (version 4.2.2; R Foundation for Statistical Computing, Vienna, Austria) using the “tidyverse” package (version 1.3.2). Individual and population predictions were plotted versus the observed data, and bias between the conditionally weighted residuals (CWRES) versus population predictions and time was evaluated. Prediction-variance-corrected visual predictive checks (pvcVPC) based on 1000 simulations were constructed to evaluate the predictive performance of the model [22]. Precision of the parameter estimates were evaluated by estimating the 95% confidence intervals (CIs) using log-likelihood profiling.
To compare the predictive performance of the final model with models from the literature, median prediction error (MdPE) and median absolute prediction error (MdAPE) were calculated according to Varvel and co-workers [23]. A MdPE lower than 20% and a MdAPE between 20 and 40% is considered clinically acceptable [24]. Two population PK models for NE were identified in literature, which are referred to as the Beloeil’s model [9] and the Oualha’s model [10]. When predicting with Beloeil’s model, we made one assumption that the covariate Simplified Acute Physiology Score (SAPS) was set to the lowest value (30) in their dataset. This was due to the fact that Beloeil’s model was built on septic shock and trauma patients and the SAPS does not extrapolate to healthy individuals.
2.4 Simulation
The final PK model was used to predict expected changes in NE PK during general anesthesia at different measured propofol concentrations. The simulation dataset was built based on the model development dataset and included all significant covariates. A typical target concentration of propofol for maintaining general anesthesia (3 mcg/mL–1) was chosen for the simulation, as well as a high concentration (6 mcg/mL–1) and absence of propofol for comparison. An infusion for 15 minutes at a dosage of 0.12 mcg/kg-1/min–1 was used for the simulation. For each patient, 1000 simulations were performed and summarized at a population level. All simulations were performed with the RxODE2 package (version 2.0.14) in R.
3 Results
A total of 36 individuals met the eligibility criteria for inclusion in the analysis. The mean age was 42 (standard deviation [SD]: 16) years, and the average BMI was 23.2 (SD: 2.4) kg/m-2. The characteristics of the volunteers are shown in Table 1. Prior to modelling, we excluded 4 outliers as they were 5-fold higher than the average concentration and affected parameter estimation. In total, 1219 samples were used in the analysis. The observed NE plasma concentration ranged from 0.25 to 67.09 nmol/L–1, with a median value of 12.90 nmol/L–1. The median value of the baseline NE plasma concentration is 2.92 nmol/L–1 (interquartile ranges [IQR]: 1.3–2.2 nmol/L–1). A summary of the baseline NE plasma concentration in the awake phase, observed median NE plasma concentration, CO and BIS at each dosing step in the general anesthesia phase is shown in Table 2. The dosing regimen and NE concentrations of some individuals in this study are shown in Fig. 1 as an example.
Table 1
Characteristics of the volunteered in the study. Data are presented as median (range) or as counts
18–34 years
35–50 years
51–70 years
n (N = sample)
12 (711)
12 (682)
12 (685)
Male/Female (N = sample)
6/6 (360/351)
6/6 (342/340)
6/6 (388/297)
Age (y)
22 (18–34)
39.5 (35–50)
57.5 (51–70)
Weight (kg)
66.9 (51.5–88.5)
72.3 (67.6–94.8)
69.6 (51.6–90.1)
Height (cm)
174.5 (160.0–196.0)
175.0 (160.0–188.0)
173.0 (151.0–192.0)
BMI (kg/m2)
22.9 (18.0–24.3)
24.4 (20.1–31.1)
23.0 (19.7–25.7)
BMI body mass index
Table 2
Summary of the NE baseline plasma concentrations in the awake phase, observed median NE concentration, CO and BIS values in the general anesthesia phase at different NE dosing steps
NE dosing step (mcg/kg–1/min–1)
Baseline
0.04
0.08
0.12
0.16
0.20
18–34 years
NE (nmol/L–1)
1.40 [1.20–2.00]
9.94 [8.72–10.55]
18.48 [15.47–21.18]
26.80 [24.82–30.41]
38.18 [33.00–40.79]
49.18 [42.31–49.78]
CO (L/min–1)
6.84 [6.10–8.73]
2.79 [2.30–3.40]
4.18 [3.40–5.56]
4.95 [4.53–6.66]
5.98 [5.41–6.69]
6.19 [5.85–6.49]
BIS
96.45 [93.48–97.33]
33.40 [27.0–36.7]
30.30 [26.90–41.80]
35.95 [30.28–44.83]
39.50 [34.10–49.03]
48.05 [42.05–52.40]
35–50 years
NE (nmol/L–1)
1.43 [1.26–2.03]
10.80 [9.78–11.57]
21.44 [18.14–22.73]
30.33 [28.92–33.40]
40.06 [37.22–44.23]
51.08 [48.41–56.56]
CO (L/min–1)
6.91 [5.66–8.01]
3.37 [2.81–3.73]
4.45 [4.04–4.65]
5.67 [4.71–6.29]
6.06 [5.66–6.47]
5.67 [5.39–5.97]
BIS
97.00 [94.73–97.45]
31.75 [25.50–35.90]
36.83 [29.38–39.00]
37.75 [35.35–40.33]
44.10 [38.90–45.98]
47.75 [45.65–49.80]
51–70 years
NE (nmol/L–1)
1.97 [1.51–2.81]
11.01 [10.13–11.25]
20.65 [19.63–21.70]
28.64 [28.19–31.35]
39.33 [34.32–42.11]
48.34 [35.73–50.72]
CO (L/min–1)
5.17 [4.63–6.25]
3.84 [3.48–4.16]
4.20 [4.08–4.44]
4.54 [4.43–5.08]
4.87 [4.60–5.13]
5.12 [4.55–5.48]
BIS
95.40 [89.70–97.50]
37.50 [36.10–40.33]
35.1 [34.83–39.40]
40.95 [39.10–45.60]
44.35 [42.08–52.34]
45.20 [42.05–51.83]
Baseline the concentration of NE before any drug administered
CO cardiac output, BIS bispectral index, NE norepinephrine
Fig. 1
Example of dosing regimen and norepinephrine (NE) concentrations in the study from individual 1–9. The blue line describes the plasma concentration of NE, the light blue area describes the dosing interval, and the vertical grey line describes the bolus of NE
×
3.1 Model Development
The two-compartmental model showed the lowest objective function value (OFV) compared to the one-compartmental model (∆OFV = 243.34 points) and the three-compartmental model (∆OFV = 252.64 points) and was selected for further model development. The addition of a lag time improved the model substantially (∆OFV = − 687.97 points) but resulted in some model instability in the estimate of the lag time. To address this, the search space for the lag time parameter was restricted to 10–16 seconds, consistent with the expected circulation time [25] necessary for NE to travel from the drug administration site to the sampling site. Removing IIV on the intercompartmental clearance parameter showed little impact (∆OFV = 1.75 points) and it was removed from the model. A combined error model was found to best describe the residual variability, exhibiting the best model fit and least bias in the goodness-of-fit plots compared to either an additive (∆OFV = 1269.35 points) or proportional (∆OFV = 65.63 points) error models.
Anzeige
Age was found to correlate with IIV CL and including an age effect on CL improved the model fit (∆OFV = − 6.5 points), in which CL decreases with age as shown below in Eq. 2:
where P denotes the PK parameter of NE, PTYP denotes the typical parameter, \({\eta }_{\text{P}}\) denotes the IIV of the corresponding parameter, \({F}_{\text{SESS}}\) denotes the session effect on the parameter. Session effect exists if the estimated value of \({F}_{\text{SESS}}\) is different from 1. We observed a session effect on the CL of NE with CL in the presence of general anesthesia being 90% (95% CI 87.3–92.0%) of that in the absence of general anesthesia. The inclusion of the session factor on CL improved the overall model fit (∆OFV = − 61.83 points) and showed improvements in the goodness-of-fit plots.
To investigate the cause of the session effect, we examined the relevant hemodynamic and anesthetic variables on the CL of NE, respectively, by introducing them as a time-varying covariates in the model with Eq. 4 described below:
where \({F}_{COV}\) represents the effect of the covariate on the CL of NE, \({\theta }_{COV\sim CL}\) was estimated to quantify the impact of the covariate on NE CL in the changes from the population median of baseline. We first explored the covariate effect of CO given that NE CL is extensively affected by overall hepatic blood flow. Furthermore, we observed a notable alteration in CO during the GA phase, with the median of CO increasing from 3.6 L/min at the beginning to 6.3 L/min by the end. In contrast, CO exhibited little variation in the awake phase, with the median of CO ranging from 6.4 L/min at the beginning to 7.0 L/min by the end. The addition of a CO as a covariate to CL improved the model fit (∆OFV = − 14.24 points) and \({F}_{SESS}\) increased from 0.90 to 0.93. We further investigated the covariate effect of measured propofol concentrations. The addition of measured propofol concentrations as a covariate to CL lowered the OFV by 60.97 points and \({F}_{\text{SESS}}\) increased from 0.90 to 1.05. Therefore, we included measured propofol concentrations as a covariate to CL instead of CO. Improvements were also seen in the goodness-of-fit plots. To evaluate the need for a session factor in the model, we removed \({F}_{SESS}\) from the model (∆OFV = 4.34 points), which did not significantly impact the overall model fit. Inter-individual variability on the covariate effect for measured propofol concentrations was identified to be statistically significant, resulting a decrease in OFV by 5.02 points upon its inclusion.
The final model included measured propofol concentrations as a time varying covariate to explain the difference in PK between the awake session and the general anesthesia session. The median measured propofol concentrations in the general anesthesia session is 3.53 mcg/mL–1, with an expected decrease in CL of 11.8%. Inter-individual variability was identified for the propofol effect on CL (∆OFV = − 5.02 points). Due to this individual variability under general anesthesia, there exists a distinction in the distribution of CL between the awake and general anesthesia phases. The distribution of CL in different phases is shown in Fig. 2.
Fig. 2
Distribution of CL in the awake phase and the general anesthesia phase. We also investigated the effect of electrical stimulation on the PK of NE by adding a factor to model to correct the individual predictions in the presence of the electrical stimulus with the equation shown below: \(C\left(t\right)=\frac{A\left(t\right)}{Vc}+C*STM\) Where \(\text{C}(\text{t})\) and A(t) denote the concentration of NE and the amount of NE at time t. C denotes the correct factor and STM denotes the presence of the electrical stimulus. The inclusion of the correction factor improved the model fit to a limited extend (∆OFV= − 4.70 points), while the removal of the propofol effect in the presence of the correct factor resulted in a substantial increase of OFV by 128.18 points. The estimated value of the correction factor is 0.66 nmol/L–1 [95% CI 0.06–1.20 nmol/L–1], suggesting that the post-stimulation NE concentration is 0.66 nmol/L–1 lower than the pre-stimulation NE concentration at a populational level. Given that the observed NE plasma concentration ranges from 6.80 to 61.04 nmol/L–1 when the stimulation is applied, a change of 0.66 nmol/L–1 was judged unlikely to cause clinically relevant effects and it was decided to remove this effect from the model
×
Anzeige
3.3 Final Model Performance
Population PK parameters of the final PK model are presented in Table 3.
Table 3
Parameter estimates of the pharmacokinetic model
Parameter
Estimate
95% CI
Lower
Upper
θ1
CL (L/min–1)
2.1
2.073
2.201
θ2
Vc (L)
2.4
1.988
2.850
θ3
Q (min–1)
0.6
0.534
0.729
θ4
Vp (L)
3.6
3.058
4.196
θ5
Prod (ng/min–1)
497.7
487.905
545.814
θ6
Lag time (s)*
13.7
13.485
34.131
θ7
CPROP~CL
− 3.57
− 4.312
− 2.780
θ8
AGE~CL
− 0.344
− 0.598
− 0.091
Parameter
IIV%CV [Shrinkage%]
95% CI
Lower
Upper
IIV_CL
10.6 [9]
7.9
14.2
IIV_Vc
44.5 [13]
33.8
63.6
IIV_Vp
40.4 [12]
32.2
57.6
IIV_prod
30.4 [8]
24.8
41.3
IIV_CPROP~CL
266.2 [32]
68.3
1515.1
δRUVProportional
16.7 [5]
15.7
17.6
δRUVadditive
32.4 [5]
27.2
38.1
Model parameters are presented as typical values along with relative standard errors (RSEs) calculated from log-likelihood profiling
δRUV residual unexplained variability, CI confidence interval, CL clearance, CV coefficient of variation [calculated as sqrt(exp(estimate)-1)×100% for interindividual variability parameters and sqrt(estimate) for within-subject variability parameters], IIV interindividual variability, Prod production rate of norepinephrine, Q intercompartmental clearance, Vc volume of distribution of the central compartment, Vp volume of distribution of the peripheral compartment
*The 95% CI of ALAG1 is not reported in our study due to the lack of sufficient information in obtaining a precise estimate of ALAG1
The goodness-of-fit plots (Fig. 3) showed that the final model adequately described the data. The population and individual predictions are in good agreement with the observed data. No obvious bias is observed over time and over population predictions.
Fig. 3
Population and individual predicted NE concentration versus observed concentrations and conditional weighted residuals (CWRES) versus population predicted NE concentrations and time after first dose. The red lines depict the locally weighted scatterplot smoothing (LOWESS) regression line
×
The pvcVPC plots of the final model presented in Fig. 4 demonstrate that the median predictions are well aligned with the observed concentrations. The peak concentrations and baseline concentrations of both sessions were well captured by the model.
Fig. 4
Prediction-variance-corrected visual predictive check (pvcVPC) of norepinephrine (NE) concentrations versus time. The red lines represent the 5th, 50th and 95th percentiles of the observed data. The shaded areas represent the 95% confidence interval of the 5th, 50th and 95th percentiles of the simulated NE concentrations
×
The predictive performance of the final model was compared with other NE models in the literature and the results are summarized in Table 4.
Table 4
Predictive performance of the final model compared with the literature models
MdAPE median absolute prediction error, MdPE median prediction error, PK pharmacokinetic
Anzeige
3.4 Literature Models
Table 4 shows a comparison of the predictive performance of our final model with NE models from the literature across different data groups. The final PK model exhibited the lowest MdPE and MdAPE for all data groups due to the fact that our model is tested for internal validity and literature models were tested for external validity. Oualha’s model and Beloeil’s model demonstrated comparable predictive performance but showed high MdPE and MdAPE for the general anesthesia data group.
3.5 Simulation
Our final model was employed to predict NE concentrations at different measured propofol concentrations (Fig. 5). The model simulations reveal that patients with a higher propofol plasma concentration are expected to exhibit higher NE concentrations at the end of the infusion, along with greater IIV, compared to patients with a lower propofol plasma concentration.
Fig. 5
Predicted norepinephrine (NE) concentrations at different measured propofol concentrations. The shaded red area displays the 95% prediction interval of the predicted NE concentrations. The black line represents the median of the predicted NE concentrations
×
4 Discussion
In this study we described the development of a population PK model of NE in healthy volunteers in both awake phase and under general anesthesia. We found that NE PK is well described by a two-compartment model with first order elimination. Notably, we observed a 10% decrease in NE CL after the induction of general anesthesia. Despite our a priori beliefs, the alterations in CO induced by general anesthesia did not provide an optimal explanation for the changes in NE PK that occur after the induction of general anesthesia. Instead, we observed that the changes in NE CL between sessions are better explained by the measured propofol concentration, and IIV in this effect exists. The effect of stimulation on NE PK under general anesthesia is very limited.
Linear elimination was employed in our model given that NE is predominantly metabolized by two enzymes that are abundant in the cell [26] and no trends indicating non-linear clearance were observed in the goodness-of-fit plots. In addition to body weight, we found that age also impacts the CL of NE. Log-likelihood profiling of the estimate describing the age effect on CL showed a narrow 95% CI with an upper CI not surpassing 0, suggesting that our study design has sufficient power for detecting a potential relationship between age and NE CL. However, the estimated age effect is rather weak, indicating patients 10 years older have about 3% lower CL. There are conflicting findings in existing literature concerning the relationship between age and NE CL, which report a decrease in CL with age, with patients 10 years older having an average lower CL of 6%. This could be attributed to the subtle nature of the age effect, which is difficult to detect [26‐31]. Nevertheless, the age effect on the PK of NE might not be clinically significant to be of practical concern.
We observed a 10% difference in NE CL between sessions and explored some potential explanations by examining relevant covariates in the model. As an intermediate-to-high extraction ratio drug [12, 13], NE CL is extensively influenced by overall hepatic blood flow. Therefore, we hypothesized that CO is the reason behind the change in CL between sessions. This hypothesis is also supported by existing literature, which has documented a correlation between NE CL and CO [27, 32, 33]. Since there was a decrease in CO during general anesthesia as a result of the combined effect of propofol and remifentanil, we suspected that this decrease of CO might lead to the decrease in CL of NE. However, including CO as a covariate in the model improved the model fit, but only to a limited extent, and \({F}_{SESS}\) still remained significant. This suggests that CO does not fully account for the changes in NE CL. A more pronounced improvement in overall model fit was obtained by including the measured concentration of propofol instead of CO. Additionally, the significance of \({F}_{SESS}\) diminished, indicating that the measured concentration of propofol effectively describes the changes in NE CL between sessions. This indicates an interaction between anesthetics and NE. Studies have been performed to investigate the effects of general anesthetics on the endogenous NE [34‐37]. Shahani and colleagues reported that several general anesthetics (e.g., propofol) have an inhibitory effect on the uptake of NE by inhibiting the NE transporters [36]. As uptake is one of the pathways by which NE is cleared out of the body, this might explain the lower CL of NE during general anesthesia. Another potential reason could be that during general anesthesia the renal sympathetic nerves activity is suppressed, which leads to a decrease in the urinary excretion of NE [38, 39].
Despite widespread clinical use of NE, its PK remains inadequately understood. There are only two other PK models reported in the literature: Oualha’s model was developed from data from critically ill children, and Beloeil’s model was developed from data including patients with critical illness secondary to sepsis and/or trauma. Both of these exhibit unsatisfactory predictive performance on the general anesthesia data group. This may be due to the fact that the covariates included (body weight in Oualha’s model and SAPS in Beloeil’s model) do not extrapolate well to general anesthesia data. Our final PK model shows better performance than the other two literature models, but this is not strong evidence because the predictions were based on the observation-fitted model. The superior performance of our final model on the general anesthesia data group can be attributed to our inclusion of the influence of the measured concentration of propofol on NE PK. Nevertheless, our model requires further clinical validation for use.
The main limitation of our study is that our model is based on data from a healthy volunteer study and is not directly translatable to certain populations with cardiovascular comorbidities or to patients who use vasoactive drugs. However, in the context of preventing IOH during the induction of general anesthesia, it is noteworthy that the healthy volunteers provide a fair representation of the general elective surgery patient population. In this population, a substantial number of patients were administered vasopressors to counteract the effects of anesthetics.
In clinical practice, NE treatment is mainly used reactively (i.e., after the occurrence of hypotension) in the form of bolus or continuous infusion to normalize blood pressure [3]. However, such empirical practice could place patients at risk of inadvertent hypo- or hypertension, which resulted from the under- or overtreatment. Previous studies have shown that even a short duration of hypotension during general anesthesia may increase the chances of developing cardiac and renal complications [40]. Optimization of NE has been attempted by automated NE administration systems based on control systems. Several research groups worldwide have been developing closed-loop vasopressor infusion systems that continuously adjust NE infusion to correct intraoperative hypotension [41‐45]. These automated systems are proven to be more efficient in maintaining blood pressure than manual administration in different types of surgeries [41, 44, 45]. Despite being automatic, these systems still function in an reactive manner and correct hypotension after it has occurred. Therefore, dosing NE in a more proactive way might be more appropriate for hypotension management [46]. A prerequisite for proactive NE administration is a better knowledge of NE PK/pharmacodynamic (PD), and the implementation of that knowledge for model-based NE administration in target-controlled infusion (TCI) or drug advisory displays. In future, more effort will be made to understand the PKPD of NE in the presence of anesthetics and concomitant opioid administration. With this knowledge, we work towards a proactive approach for patient groups at higher risk of developing IOH and provide them with evidence-based hemodynamic management.
In conclusion, The NE PK is well described with a two-compartment model with a first-order elimination. The NE CL exhibited a 10% decrease under general anesthesia, with this difference being attributed to the measured concentration of propofol. An age effect was identified on NE CL; however it might be too subtle to be clinically significant. The impact of stimulation on NE PK under general anesthesia is very small and might not be of practical concern.
Declarations
Funding
The study is supported by the Department of Anesthesiology at University Medical Center Groningen, The Netherlands.
Conflicts of Interest
M.M.R.F Struys: His research group/department received (over the last 3 years) research grants and consultancy fees from The Medicines Company (Parsippany, NJ, USA), Masimo (Irvine, CA, USA), Becton Dickinson (Eysins, Switzerland), Fresenius (Bad Homburg, Germany), Dräger (Lübeck, Germany), Paion (Aachen, Germany), Medtronic (Dublin, Ireland). He receives royalties on intellectual property from Demed Medical (Sinaai, Belgium) and the Ghent University (Gent, Belgium). He is an editorial board member and Director for the British Journal of Anaesthesia. D.J. Eleveld: Has received travel from Becton Dickenson and is an editorial board member of Anesthesiology journal. P.J. Colin: Over the last 3 years his research group has been involved in contract research for PAION UK Ltd. (London, England) and Acacia Pharma Ltd. (Cambridge, England). Yingxue Li, Jeroen V. Koomen, Johannes P. van den Berg, Jaap Jan Vos, Ilonka N. de Keijzer declare that they have no conflicts of interest that are relevant to the content of this article.
Ethics Approval
The protocol of the study has been approved by the local medical ethics committee.
Consent to Participate
Informed consent was acquired from each individual participant incorporated in the study.
Consent for Publication
Not applicable.
Availability of Data and Material
The data and material are accessible upon reasonable request by contacting the corresponding author.
Code Availability
The model code is accessible upon reasonable request by contacting the corresponding author.
Authors’ Contributions
M.M.R.F. Struys, P.J. Colin, J.P. Van den Berg, J.J. Vos: conception and design of the study and manuscript revision; J.P. Van den Berg, J.J. Vos, I.N. de Keijzer: data collection, bioanalysis and manuscript revision; Y Li, J.V. Koomen, D.J. Eleveld, P.J. Colin: data analysis and interpretation, modelling, manuscription preparation and revision.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License, which permits any non-commercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc/4.0/.
