Predictive Accuracy of Target-controlled Propofol and Sufentanil Coinfusion in Long-lasting Surgery

This study was approved by the Institutional Committee of Human Ethics. Informed consent was obtained from 10 nonobese adult patients with American Society of Anesthesiologists status I and II (seven men and three women) and age 36–69 yr. None of the patients had a history of cardiovascular, respiratory, neurologic, hepatic, or kidney dysfunction. All of the patients underwent the same prolonged type of cervicofacial surgery for extended tumor and multiple lymph node resection. This population included six pharyngo-laryngectomies, two transmandibular buccopharyngectomies, one orbital tumor with hemifacial resection, and one tumor of the acoustic meatus with partial temporal bone resection.

The patients were premedicated with intramuscular midazolam (0.1 mg/kg) 30–60 min before their arrival in the operating room. After noninvasive monitor placement, the following blood vessels were cannulated: a forearm vein for administration of anesthetics and other drugs; a femoral vein for fluid administration, blood, or plasma replacement; a femoral artery for the continuous monitoring of blood pressure and blood sampling; and a subclavian vein for monitoring the central venous pressure. Hartmann solution was infused at a rate of 80 ml/h in addition to blood and fluid replacement as indicated by the clinical context. All fluids were warmed to 37–40°C.

Heart rate, mean arterial pressure, and mean central venous pressure were noted before induction and strictly repeated on a regular basis thereafter.

Patients received total intravenous anesthesia with a combination of propofol and sufentanil TCI using the pharmacokinetic sets of Marsh et al.  6and Gepts et al. , 7respectively (table 1). To control the propofol and sufentanil TCI, we used Toolbox, the software elaborated in the Department of Computer Science of the Erasmus Medicine School of the Free University of Brussels. 8Two Pilot C pumps (Becton Dickinson Inc., Franklin Lakes, NJ) were driven through the two serial ports of a portable personal computer.

Sufentanil infusion was started to produce within 60 s a 0.5-ng/ml plasma concentration. One minute later, the propofol infusion was initiated to achieve a plasma target propofol concentration of 3 μg/ml; this was increased to 4 μg/ml when needed to achieve unconsciousness. A bolus dose of atracurium 0.5 mg/kg was then administered intravenously to facilitate tracheal intubation. No additional atracurium was given thereafter. After intubation, patients were ventilated with an air–oxygen mixture (40% oxygen), with maintenance of end-tidal carbon dioxide pressure between 32 and 35 mmHg.

Anesthesia management was standardized according to decisional algorithms. The volumes of peroperative and postoperative blood losses were noted. When hypovolemia occurred, 10–25 ml/kg serum albumin 4% was used as colloid to maintain mean arterial pressure and mean central venous pressure within acceptable values. If hemoglobin blood measurement (Instrumentation Laboratory co-oximeter; Lexington, MA) was less than 90 g/l, 10–20 ml/kg of packed erythrocytes was transfused to the patient to restore the selected threshold level of hemoglobin.

Arterial blood samples were drawn every 5 min for 20 min, every 10 min for the next 40 min, and hourly up to the end of the surgery. When propofol or sufentanil target concentrations were changed, additional samples were drawn, approximately 2–3 min after the change. With the TCI pump delivering at the maximum speed of 15 ml/min, we expected the plasma concentration to have reached the new steady state.

For propofol, 3 ml of blood was allowed to clot in a tube prepared with silicon and was placed in a refrigerator (4°C) until analysis. The measurements of propofol concentrations were performed by high-performance liquid chromatography with fluorescence detection. 9The limit of detection was 0.001 μg/ml, and the range of quantification was 0.002–10 μg/ml.

For sufentanil, blood samples (4 ml) were immediately centrifuged in the operating room, and separated plasma was frozen (−27°C) for storage until time of analysis. Plasma sufentanil concentrations were determined by radioimmunoassay 10(Janssen Research Foundation, Beerse, Belgium). The limit of detection was 0.02 ng/ml and the interassay coefficient of variation was 8.5–10.5% for a concentration range of 0.05–10 ng/ml.

Data were analyzed using SyStat 5.0 (Systat Incorporation; Hallogram Publishing Inc., Aurora, CO). We used descriptive statistics, including means, 95% confidence intervals, medians, and 10th and 90th percentiles. For each sample, the percentage performance error (PE) of the predicted propofol and sufentanil plasma concentrations were calculated according to the recommendations 3,11:

PE is an indication of the bias of the achieved concentrations, and the absolute value of PE (|PE|) is a measure of the precision (inaccuracy).

Intrasubject data analysis consisted in an evaluation, according to Coetzee et al.  3and Varvel et al.  11of four indicators of predictive performance for the n subjects.

The percentage median prediction error (MDPE) reflects the bias of the TCI in the ith subject.

MDPEⁱ= median {PE^ij, j = 1,…, Nⁱ}

where Nⁱis the number of PE values obtained for the ith subject.

The percentage median absolute prediction error (MDAPE) indicates the inaccuracy of TCI in the ith subject.

MDAPEⁱ= median {|PE|^ij, j = 1,…, Nⁱ}

The divergence 3,11for the ith individual is defined as the slope of the linear regression equation of the absolute value of the percentage PE (|PE|) against time expressed in units of percentage divergence per hour. A positive value indicates an increasing widening of the gap between predicted and measured concentrations, whereas a negative value conveys a convergence of the measured to the predicted values.

Wobble 3,11represents an index of the time-related changes in performance and measures the intrasubject variability in PEs. In the ith subject the percentage wobble is calculated as follows:

wobbleⁱ= median {|PE^ij− MDPEⁱ|, j = 1,…, Nⁱ}

Figures 1–5show the performances of each model over time. Figure 1shows measured values divided by the predicted values plotted over time. These graphs provided a means to visually assess the extent of bias and precision in both groups. Figure 2shows plots of the regression lines for PE^ij− MDPEⁱon time and illustrates how the errors in each patient tended to differ over time from that patient’s own MDPEⁱ. Figure 3shows plots of (PE^ij− MDPEⁱ) over time (where j = 1,… Nⁱ in the ith subject) and illustrates the magnitude of the terms used to estimate the wobbleⁱof each patient. It is the median of these absolute values that provides the wobbleⁱof each patient. Once each patient’s wobble has been obtained, the wobble of the model is derived by calculating their average. Figure 4shows patient data with the most prolonged infusion time (1,020 min), providing the infusion rate and the target and measured concentrations for propofol and sufentanil. Figure 5shows the results from the individual patients exhibiting the best, median, and worst performances, illustrating the adequacy of both models (as judged by the measured/predicted ratio, PE-MDPE, and divergence trends).

Demographic data are summarized in table 2. The ranges of applied target concentrations was 2– 5 μg/ml for propofol and 0.2–1 ng/ml for sufentanil. Table 3shows the results of the four calculated indicators (MDPE, MDAPE, divergence, and wobble) of predictive performance for the pharmacokinetic sets during a mean period of 762 min. MDPE and divergence did not differ statistically from zero as their 95% confidence intervals include 0.

Figure 1shows that there was no systematic overprediction of the concentrations of both drugs, especially during the initial 60 min, and in some patients the sufentanil model underpredicted the measured concentrations. Figure 2illustrates that the tendency for the error to range in the negative values and from ith patient’s own MDPEⁱis comparable and similar for both drugs. The wobble analysis (fig. 3) showed a similar degree of variation from each patient’s MDPEⁱover time for the two drugs. Figure 4allows the comparison between the calculated and measured plasma concentrations in parallel with the infusion rate recorded on the pump. Figure 5is a two-panel plot (one for each drug) from the patients presenting the best, median, and worst performance of the propofol and the sufentanil pharmacokinetic models. These patients were selected on the basis of their measured/predicted ratio (fig. 1), divergence (fig. 2), and PE-MDPEⁱ(fig. 3). The target concentrations for propofol and sufentanil were modified two, two, and three times, respectively, for propofol and four, six, and one times, respectively for sufentanil in each patient.

Finally, in agreement with the anesthesia management standardization, the patients received, on average, 6.2 ml · kg⁻¹· h⁻¹(range, 6–6.6 ml · kg⁻¹· h⁻¹) of crystalloid (Hartmann solution), 4.1 ml · kg⁻¹· h⁻¹(range, 4–4.3 ml · kg⁻¹· h⁻¹) of colloid (serum albumin 4%), and 1.9 ml · kg⁻¹· h⁻¹(range, 1.8–2.1 ml · kg⁻¹· h⁻¹) of packed erythrocytes, respectively. All of the subjects were transfused at least one time to maintain hemoglobin more than 90 g/l during surgery.

Because infusion was long, drug accumulation on time could be suspect. Probably, in relation to the relatively small range of measured concentrations (2–5 μg/ml for propofol and 0.2–2 ng/ml for sufentanil), the variance analysis shows that the errors distribution seemed to be dose-independent. This rejects any possible relation between the prediction errors and the dose of the delivered drug.

The study was designed to investigate prospectively the pharmacokinetics of long-term TCI coadministration of propofol and sufentanil in strictly standardization clinical conditions. It was also designed to interfere as little as possible with the pharmacokinetic of the drugs. For this we considered (1) the selection of a surgery interfering poorly with drug metabolism and pharmacokinetics and (2) the anesthesia management standardization and the guidelines dealing with the replacement of blood volume losses in agreement with the present consensual rules. 12,13 

Several models have been proposed and validated for their ability to predict drug concentration in the plasma compartment for short-term studies. Our main goal was to validate those in such a long-term study, this having not been performed previously for propofol and sufentanil. We expected the same order of magnitude of the error or even greater. We would not have been surprised to observe increasing divergence over time between measured and predicted concentrations. Of course, we suspected that the use of the decisional algorithm had contributed to the patient’s homeostasis and so had contributed to the reduction of percentage prediction error on a sample measurement for a given patient. Compartment pharmacokinetic models are useful to predict theoretical drug concentrations in the central compartment 14when the targeted plasma concentrations are restricted to values less than 5 μg/ml for propofol 15and 1 ng/ml for sufentanil. 7The pharmacokinetic parameter sets were selected from the literature. For propofol, according to Coetzee et al. , 3the modification by Marsh et al.  6of the model by Gepts et al.  16was selected. For sufentanil, the multidose model of Gepts et al.  7was chosen. In comparison to the results reported by Coetzee et al. , 3arterial propofol concentrations revealed greater bias (MDPE −12.1% in comparison to −7%) but is not statistically significant. The results showed slightly greater inaccuracy (MDAPE 22.1% instead of 18.2%), whereas divergence is smaller (−2.92 vs.  6.5%) and also not statistically significant. Finally, wobble is of the same order of magnitude (11.6 vs.  10%). Because results with regard to sufentanil were lacking, we simply obtained results as good as those obtained for propofol (MDPE −10%, MDAPE 20.7%), except for wobble, which is the worst (22.3% instead of 10%), and divergence, which is even better (−0.22% compared with 6.5%) and also not statistically significant.

There is no agreement in the literature about the corrections of the pharmacokinetic parameters for age and sex. For propofol, several investigators 14,17,18reported influence of age or sex on the parameters, whereas the results of others 19–21are not as clear. Results for sex and age simply do not exist for sufentanil presently. Pharmacokinetic interactions between propofol and opioids (fentanyl, alfentanil) have been regularly studied 22–26using an opioid bolus dose 22,24,25or infusion. 26,27These results are not homogeneous as significant interactions were described 22,23,26concerning the propofol pharmacokinetic parameters when alfentanil is coadministered. On the contrary, the coadministration of propofol and fentanyl, in the conditions described by Gill et al. , 24are not accompanied by an alteration in the propofol pharmacokinetic model.

All patients underwent the same prolonged cervicofacial surgery. This kind of procedure does not induce major pharmacokinetic perturbations 1,27or extremely painful stimuli. 27–31Moreover, surface or nonmajor surgery required, in general, lower anesthetic drugs concentrations. 21,32Lengthy surgical procedures do not automatically imply difficult changes to manage in the macrophysiologic parameters (cardiovascular, hepatic, and renal) controlled. Shafer et al.  21and Björkman et al.  33demonstrated that no substantial interference with the pharmacokinetics of propofol and opioids was observed when cardiac output was maintained in a physiologic range of values. This would correspond to the conditions in the present study and can be considered as acceptable. Furthermore, it was important to obtain usual, and not too-high, propofol target concentrations during the study, because the quality of prediction of different pharmacokinetic models is poorer for high values of blood concentrations. 15Blood losses were essentially external throughout the surgical field and therefore more simple to observe and quantify.

It must be emphasized that, in this such topographically restricted surgery, the prevention of relevant heat loss and plasma hypovolemia seemed more simple to manage compared with plastic surgery or peripheral vascular surgery, where large skin areas are frequently exposed to room temperature.

In conclusion, the selected pharmacokinetic parameter sets, those of Marsh et al.  6for propofol and Gepts et al.  7for sufentanil, proved to be accurate for predicting plasma concentrations during prolonged surgery (6–17 h) with an acceptable MDPE, wobble, and divergence and without any important time influence. These results represent a major point of the study, particularly for the sufentanil TCI, because these notions were until now lacking in the literature. Finally, the models could be used to predict propofol and sufentanil blood concentrations, providing that standardized rules for the management of TCI and volume replacement could be applied.

The authors thank Zeneca Pharmaceuticals Ltd., Destelbergen, Belgium, for the measurements of propofol by high-pressure liquid chromatography, and Dr. Woestenborghs, Janssen Pharmaceuticals Ltd., Beerse, Belgium, for the measurements of sufentanil by radioimmunoassay.