| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
| ||||||||||||||
| ||||||||||||||
|
|
|
|
||||||||||||
|
||||||||||||||
*Department of Anesthesiology, University Medical Center Groningen, The Netherlands; Department of Anesthesiology and Intensive Care Medicine, University of Bonn, Germany; Medical Department, Division Hospital Care, B. Braun Melsungen AG, Melsungen, Germany; Department of Anesthesiology, Lüneburg, Germany
Address correspondence and reprint requests to J. K. Götz Wietasch, MD, Department of Anesthesiology, Groningen University Hospital, PO Box 30001, 9700 RB, Groningen, The Netherlands. Address e-mail to j.k.g.wietasch{at}anest.umcg.nl.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
TCI administration of propofol is often used in combination with opioids, some of which alter propofol kinetics (2,3). Remifentanil appears to be an ideal analgesic component for total IV anesthesia (TIVA) in combination with propofol because of its elimination via an independent pathway from that of propofol as well as its rapid elimination and favorable controllability. However, there have been inconsistent findings concerning a propofol and remifentanil interaction in the literature (4–6). The aim of this study was to investigate the performance of propofol TCI delivery when combined with remifentanil in patients undergoing elective surgery.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Preoperative medication included flunitrazepam 1 mg and ranitidine 150 mg orally the evening before surgery and midazolam 7.5 mg and ranitidine 150 mg orally 1 h before anesthesia. On arrival in the operating room, electrocardiogram leads, pulse oximetry, and 4 frontal electroencephalogram (EEG) leads (bi-referential placing; A-1000, V. 3.12; Aspect® Medical Systems International B.V., Leiden, the Netherlands) were attached. After insertion of an IV and an arterial catheter (radial artery), anesthesia was induced with propofol TCI using an initial plasma target concentration of 5 µg/mL. After loss of consciousness, cisatracurium (0.15 mg/kg) was given for muscle relaxation. Remifentanil TCI was started with an initial plasma target of 5 ng/mL. After intubation of the trachea, the lungs of the patients were mechanically ventilated (oxygen/air, Fio2 = 0.4). Propofol target concentrations were adjusted to maintain BIS (bispectral index) values of 50 ± 10. Remifentanil target concentrations were adjusted to heart rate and arterial blood pressure. A change of more than 30% from baseline values, if not related to blood loss, indicated an adjustment of the target concentration in the same direction. Arterial blood pressure, heart rate, oxygen saturation, and multi-parameter EEG (95% spectral edge frequency [SEF95], BIS) were recorded digitally, as well as the TCI dosing histories of propofol and remifentanil.
Arterial blood samples for determination of propofol plasma concentrations were taken at baseline, loss of consciousness (time range: 1.1–4.5 min), tracheal intubation (time range: 5.0–11.9 min), 15, 30, and, if possible (depending on time until begin of surgery), 60 min after start of propofol TCI but before surgery, at skin incision and 30, 60, and, if possible (depending on the duration of the surgery), 120, 180, 240, 300 min after skin incision. Additional blood samples were taken before stopping propofol TCI and at tracheal extubation. Blood samples were immediately centrifuged (2500g at 3°C) and plasma supernatants were kept frozen at –80°C. Propofol plasma concentrations were determined in duplicate by high pressure liquid chromatography according to Plummer (12). One sample was analyzed in our analytic laboratory (limit of quantification = 0.05 µg/mL, intraday precision < 6.1%, interday precision < 3.2%, linear range tested in human plasma 0.05 – 10 µg/mL); the other sample was analyzed in a commercial reference laboratory (IKP, Grünstadt, Germany; limit of quantification= 0.05 µg/mL; intraday precision < 9.8%, interday precision < 9.9%, linear range tested in human plasma 0.14 – 11.11 µg/mL).
We quantified the performance of the TCI system using the median absolute performance error (MdAPE) and the median prediction error (MdPE) according to Varvel et al. (13). The performance error (PE) calculation was as follows:
|
|
where Cm is the measured plasma concentration and Cp is the predicted (calculated) plasma concentration, calculated using the dosing history and the Marsh et al. (8) or Schnider et al. (14) variable set. The accuracy of the TCI System for the i th patient is given by the MdAPE:
|
|
where Ni is the number of samples obtained for the ith patient. The bias of the TCI system for the i th patient is given by the MdPE:
|
|
In the current investigation the PEs are known in some individuals with more certainty than in others because of varying number of samples (j = 9 to 15). The simple pooled data approach was applied, which has the advantage of weighting the measured values of each individual to calculate MdAPE and MdPE for a typical patient. However, this approach allots a patient from whom many samples are drawn more influence than a patient from whom few samples are drawn.
|
|
|
|
where M is the number of investigated patients.
Pharmacokinetic analysis was performed with NONMEM V (NONMEM project group, University of California, San Francisco) using a mixed-effect population model to estimate both intraindividual and interindividual variability for mammillary 2- and 3-compartment models using a proportional error model. The compartmental model was described by the following parameters: Vc = central volume of distribution, CL= elimination clearance, V2 and V3 = fast and slow compartment, and Q2 and Q3 = fast and slow transfer rates. The population parameters were characterized by the standard error of estimates, and the interindividual error (coefficient of variance, exponential model).
The very large number of changes in propofol infusion rate (up to 2848 per individual) made a straightforward NONMEM analysis prohibitively computationally intensive. To facilitate NONMEM analysis, we reduced the size of the individual data sets using a calculated virtual mean infusion rate that changes only if a predefined infusion threshold (
VA) was exceeded (15). We found that a VA = 6.4 mg/min reduced the data set to a maximum of 85 entries per patient with little influence on the estimated pharmacokinetic variables. The data reduction algorithm was implemented in a Visual basic routine for Microsoft Excel® (V. 97; Microsoft).
The influence of covariates and different propofol preparations was investigated in all 54 patients using a noncompartmental analysis. CL, elimination rate constant (kel), mean residence time (MRT), and steady-state apparent volume (Vss) were derived from the total propofol dose and the area under the concentration time curve extrapolated mono-exponentially to infinity. Drug elimination was assumed to occur from the central compartment with first-order kinetics. The influence of type of surgery (abdominal versus nonabdominal), gender, group, physical status, and smoker status on propofol pharmacokinetics were assessed by comparison of the area under the concentration time curve. The area under the concentration time curve was calculated from start (T0) to end of propofol application (T14) using the trapezoid rule. To compare individuals with varying times of propofol administrations, we normalized the area under the curve (AUC) to the duration of anesthesia (AUCI0-t).
We also analyzed the propofol data with a 3-compartment model with (full model) and without (reduced model) the type of propofol preparation as a covariate. According to the likelihood ratio test, a decrease of the objective function of more then 3.84 (
2, 1 degree of freedom, 
-level 0.05) was considered to be significant.
To identify the influence of additional covariates, regression analysis was performed of the estimated individual pharmacokinetic variables against weight, height, age, gender, ASA status, and type of surgery.
The primary goal of the study was to investigate the performance of the propofol TCI system in combination with remifentanil. This question was investigated in a two-step analysis. In a first step, the performance of the propofol TCI system and the pharmacokinetic variable set of propofol were analyzed in all patients using the data reduced propofol dosing history and the measured and predicted propofol plasma concentrations (described above).
In a second step, we performed a post hoc analysis of the data after completion of the study. Two groups of 27 consecutive patients were built and analyzed separately. In the first set of 27 patients (Group 1) a new 3-compartment variable set for propofol was determined. This new variable set was evaluated together with the Marsh et al. (8) and Schnider et al. (14) variable set in the second set of 27 patients (Group 2). The predicted propofol plasma concentrations were recalculated from the reduced dosing histories and the corresponding variable set. Performance was calculated as described above.
The data are expressed as mean ± sd (range). Biometric data were compared with the Student's t-test. Differences between non-compartmental parameters for analysis of covariates were assessed by the Mann-Whitney U-test with Bonferroni correction for 8 repeated comparisons. P < 0.05 was considered significant.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
|
Mean total infusion time was 230 ± 86 min (range, 61–487 min), and mean total doses of 1951 ± 645 mg propofol (range, 300–3885 mg) and 3824 ± 2543 µg remifentanil (range, 1064–16019 µg) were administered. The mean measured plasma concentration of propofol was 5.8 ± 1.3 µg/mL and deviated significantly from 3.3 ± 0.4 µg/mL predicted by the TCI system. The propofol TCI system (Marsh et al. [8] variable set) showed a poor performance with MdPE of 58.6% and MdAPE of 60.7%. Recalculation of the PE using the Schnider et al. (14) variable also showed a poor performance, with MdPE of 47.9% and MdAPE of 50.3%.
The noncompartmental pharamcokinetic analysis provided the following variables: MRT = 102 ± 35 min; Vss = 2.27 ± 0.98 L/kg; kel = 2.82 ± 1.62 L/h; CL = 22 ± 5 mL/min/kg. The extrapolated terminal AUC was small at <5% of the total AUC.
All patients' plasma propofol concentrations could be satisfactorily described by a mammillary two compartment model (Table 2). A first attempt to fit the dosing history and the measured plasma concentrations to a mammillary 3-compartment model was numerically unstable. To address this we used the volume of distribution (Vss) found in the noncom-partimental analysis to set the volume of the slow compartment to V3 = VssN V1 – V2. This approach was numerically stable and resulted in an objective function of 904.9, significantly less than from the 2-compartment model, indicating that the 3-compartment model was more suitable. We found a central volume of distribution (Vc) of 3.55 L and CL of 1.31 L/min (Table 2).
|
Neither AUCI0-t nor CL were significantly different between abdominal versus nonabdominal surgery, propofol preparation, or nicotine abuse. We observed a lower AUCI0-t (P < 0.05) and a slight increase (not significant) in CL in female patients. In ASA III patients CL was significantly lower in comparison with ASA II patients (Table 3). Propofol preparation as a covariate did not alter the objective function of the pooled data (full model = 901.5 versus reduced model = 904.9) and similar pharmacokinetic variables were estimated. Therefore, propofol preparations were considered for this study to be clinically bioequivalent.
|
None of the covariates tested (gender, age, weight, height) reached significance in the parameter-covariate relationship. Moreover, neither the kind of surgery nor the duration of infusion showed an influence on MdPE or MdAPE (data not shown).
On the post hoc analysis of the data there were no differences in biometric data between the first 27 patients (Group 1) and the subsequent 27 patients (Group 2), or between patients receiving propofol LCT or propofol MCT/LCT (13/14 in Group 1 versus 14/13 in Group 2). There were also no differences between groups regarding the time of anesthesia, type of surgery, measured or predicted propofol plasma concentration, or depth of anesthesia quantified with the SEF95 or BIS (Table 1).
In the second step of the study, compartmental analysis of Group 1 achieved a good fit (MdPE = –1.1%, MdAPE = 21.4%) and resulted in pharmacokinetic variables close to those found for the entire group (Vc = 3.16 L, V2 = 10.0 L, V3 = 160 L, CL = 1.33 L/min, CL2 = 1.35 L/min, CL3 = 0.64 L/min). Recalculation of PEs from the dosing history in Group 2 showed a good performance for the new propofol variable set (MdPE = 2.4%, MdAPE = 21.2%). Initial propofol plasma concentrations (T1) were still slightly underpredicted. The reason for this is unclear but may have been a result of model misspecification. Predictive performance of the new propofol variable set was better than that found using Marsh et al.'s variable set (8) (MdPE= 57.3%, MdAPE = 58.5%) or Schnider et al.'s variable set (14) (MdPE= 49.5%, MdAPE = 50.4%) (Fig. 1).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Previous studies using the Marsh et al. variable set (8) also found a systematic underestimation of plasma propofol concentrations (1,3,4,10,16). Despite this inaccuracy, the reliability and safety of the Marsh et al. variable set has been shown in randomized controlled trials in routine patients (11,17). Other studies have indicated that the Schnider et al. variable set (14) was more robust in the prediction of propofol concentrations (18); however, we found only a minor improvement in performance using this set.
To investigate the misprediction of propofol TCI we performed a noncompartmental analysis of propofol pharmacokinetics. This resulted in a Vss of 2.3 L/kg, a value comparable to other investigations (8,14,19), whereas our CL of 22 mL/min/kg is less than described by others. Two- and 3-compartmental pharmacokinetic analysis of the propofol data provided more information about possible changes in propofol pharmacokinetics. Our major findings were: 1) a smaller Vc, 2) a smaller Q2, and 3) a lower CL compared with the variables described by Marsh et al. (8) (Table 2).
The discrepancy of the smaller Vc and the smaller Q2 found in this investigation in comparison with the study of Marsh et al. (8) may be attributable to differences in sampling site (10) and scheme (20). Because Marsh et al. (8) obtained venous sampling mostly under steady-state conditions, the value of the Vc is restricted to larger values only (i.e., Vc = 17.1 L and Q2 = 1.92 L/min). Rapid sampling after a single bolus injection as performed by Schnider et al. (14) covered the first distribution phase of propofol and provided smaller values for Vc and Q2 (4.3 L and 1.29 L/min, respectively). This approach has the drawback of incomplete mixing effects (21) and model misspecification (14) that might lead to an underestimation of these variables. We choose a slow bolus infusion scheme (TCI with propofol maximum infusion rate of 133 mg/min) in combination with early (1.1–4.5 min) arterial blood samples, which should be nearly optimal for determining Vc (20,22). This method leads to values of Vc of 3.6 L and Q2 of 0.98 L/min (Table 2). Interestingly, the Vc found here matches well with the circulating blood volume found in indocyanine green investigations (21).
The relatively small CL found in this study (CL = 1.3 L/min; Table 2) may be a result of metabolic and pharmacodynamic influences, as differences in sampling scheme should not influence CL (20). Remifentanil is eliminated by unspecific plasma esterases, whereas propofol is glucuronidated and degenerated by cytochrome P450, and, therefore, direct metabolic interaction seems unlikely. An investigation from Bouillon et al. (6) in healthy volunteers did not show any influences of remifentanil on propofol performance; however, there are some inconsistencies in their pharmacokinetic analysis. They found a propofol CL of 3 L/min that is not only 50% larger than known from other investigations (8,10,14) but also larger than the liver blood flow (23), which we consider to be unrealistic.
Our findings are consistent with results of other clinical investigations (4,5) that also found an underestimation of propofol plasma concentrations ranging from 17% to 60%. A possible explanation of this difference might be an indirect pharmacodynamic effect on pharmacokinetics. Propofol exerts profound centrally mediated sympatholytic effects (24,25) associated with a decrease of cardiac output (26) and hepatic blood flow (23). Mertens et al. (27) demonstrated a hemodynamic interaction mechanism for propofol and alfentanil. Such an interaction with central effects of propofol might also be expected with remifentanil (28). This effect might be more pronounced in patients than in healthy volunteers and is compatible with the significant reduction in CL found in our ASA III patients (Table 3). In the present study, the pharmacodynamic effect of propofol/remifentanil coadministration was not a target variable of the investigation. It can therefore only be speculated that a pharmacodynamic synergistic interaction of these drugs on sympathetic drive, cardiac output, and hepatic blood flow could be the mechanism of pharmacokinetic interaction. Further studies would be required to investigate this hypothesis in more detail.
Investigating pharmacokinetics under routine clinical circumstances has disadvantages that may impact the pharmacokinetic variables obtained. The dosing schemes are determined by the clinical requirements and are thus not always optimally designed for pharmacokinetic analysis. The physical status of patients is usually impaired, and the sympathetic depression by anesthetics pronounced, in comparison to healthy volunteers. Also, surgery is associated with painful stimuli and sometimes with severe disturbance of patient homeostasis (i.e., volume shift, blood loss, or endocrine activation). However, because pharmacokinetic variables are more likely to describe pharmacokinetics within the individual study design (1), clinically obtained variable sets seem to be a better choice for accurate prediction of plasma concentrations in clinical practice.
A limitation of our pharmacokinetic analysis is the short postinfusion sample scheme. In the noncompartmental analysis the extrapolated terminal AUC may be underestimated and thus our estimated Vss and CL may be too large. However, after a mean total infusion time of propofol of nearly 4 hours, the extrapolated AUC was <5% of the total AUC and therefore the influence on Vss and CL should be limited.
The sampling scheme was not designed for identification of V3 in the compartmental analysis, as more late samples after termination of propofol infusion would have been required. This leads to numerical instability for a standard 3-compartment analysis. We solved this problem by using V3 calculated from the noncompartmental analysis as a priori information (29).
The clinical implications of our investigation are that the use of either Marsh et al.'s (8) or Schnider et al.'s (14) variable set leads to underestimation of propofol plasma concentrations during induction and maintenance of anesthesia (Fig. 1). The use of the new variable set determined in part 2 of this study may allow improved performance, as evidenced by the more accurate predictions shown in Figure 1. None of the sets tested could accurately predict all propofol plasma concentrations at T1; this may have been a result of insufficient mixing or other model misspecification problems.
To depict the clinical effects of our findings, we simulated a propofol TCI administration using the variable sets of Marsh et al., Schnider et al., and our own for a plasma target value of 5 µg/mL and recalculated the plasma concentrations using the new established variable set (Fig. 2). Marsh et al.'s variable set leads to a considerable overshoot of propofol plasma concentration during the initial phase of TCI application. Despite the poor prediction accuracy in this case, this kind of initial overshoot may offer certain practical advantages under clinical conditions. Initial hypnotic onset occurs faster than if the Schnider et al. or our set had been used, although the risk of hemodynamic instability might be increased. At the same time, the systematic error of Marsh et al.'s and Schnider et al.'s variable sets may lead to excessive dosing if TCI is used as the single means to achieve adequate hypnosis and this may cause a delayed recovery.
|
The authors are grateful to Klaus Retzman for propofol analysis and to Jean-Louis Griffoul for implementation of the pharmacokinetic data sets to the TCI systems. We express our thanks to Dr. Thomas Bouillon, Bern University Clinic, Switzerland, and Dr. Karel Kuizenga and Dr. Douglas Eleveld, Groningen University Clinic, The Netherlands, for their helpful suggestions in the preparation of this manuscript.
|
Footnotes |
|---|
Accepted for publication August 31, 2005.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  A. Rigouzzo, L. Girault, N. Louvet, F. Servin, T. De-Smet, V. Piat, R. Seeman, I. Murat, and I. Constant The Relationship Between Bispectral Index and Propofol During Target-Controlled Infusion Anesthesia: A Comparative Study Between Children and Young Adults Anesth. Analg., April 1, 2008; 106(4): 1109 - 1116. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
