In Reply:—
Pharmacokinetic modeling has gained more than academic interest because these models can be used for computer-controlled drug administration, which probably will become more and more relevant for clinical practice. This aspect of “applied pharmacokinetics” was one reason for the population pharmacokinetic analysis of propofol published in Anesthesiology, 1and, obviously, it also was the reason for the criticism of Vuyk et al. as pointed out in their letter. Because discussion is essential for scientific progress, we appreciate any comment about our work, but we do not believe that the arguments of the authors can justify their conclusions.
Vuyk et al. claim that the pharmacokinetic parameter set published in Anesthesiology 1is “unsuitable for application in TCI [target-controlled infusion] in elderly patients.” This statement is, however, nothing but an unproved claim because Vuyk et al. did not investigate, either retrospectively or prospectively, the predictability of different parameter sets but simply compared the results of different studies. The cited publications of Dyck and Shafer, 2Schnider et al. , 3and Oostwouder et al. 4also present estimates of the pharmacokinetic parameters of propofol without any further validation. Therefore, we completely agree with the statement that “nobody knows … whether this parameter set predicts the measured propofol concentrations better than previous parameters sets”—nobody, including Dr. Vuyk et al.
Therefore, the problem is reduced to the question of the comparability of different data sets: In our population analysis, we studied 270 individuals, with 35 subjects aged 65 yr or older. From these 35 individuals, 9 were administered propofol as a single bolus dose. Thus, there were 26 individuals aged 65 yr or older who were administered propofol as a continuous infusion. This is only 10% of the complete data set, but the absolute number of individuals is still greater than in the study of Schnider et al. 3(9 elderly volunteers of 24 in toto ) and comparable with the studies of Oostwouder et al. 4(22 elderly) and Dyck and Shafer 2(20 elderly). Regarding the relatively small fraction of elderly patients in our total population, it should be noted that the effect of age on elimination clearance was highly significant (age as a covariate for clearance led to a significant reduction of the NONMEM 5objective function) and distinct (reduction by approximately 50% for a patient aged 75 yr). If the number of elderly individuals within the total population had been too small, the effect of age should have been much smaller than observed.
When comparing the pharmacokinetic parameters of our study with those from Dyck and Shafer, 2Oostwouder et al. , 4and Schnider et al. , 3the differences in the estimates of the elimination clearance and the central volume of distribution are obvious. We found relatively small values for clearance (0.9 l/min for a 75-kg patient aged 75 yr) compared with Dyck and Shafer, 2Oostwouder et al. , 4and Schnider et al. 3(approximately 1.7 l/min). Vuyk et al. claim that this is a result of the short sampling period in our data. For short sampling periods, however, the opposite should be observed. Because the distribution into deep peripheral compartments can not be identified under these circumstances, the model should overestimate the elimination clearance.
Regarding the central volume of distribution, Oostwouder et al. , 4Dyck and Shafer, 2and Schnider et al. 3found extremely small values of approximately 4–6 l, compared with our estimates of 9 l for young adults and 7 l for a 70-yr-old individual, when propofol is administered as a continuous infusion. These differences demonstrate again a widely discussed methodologic problem in pharmacokinetic data analysis. The estimation of central volume of distribution depends on the sampling procedure in the early phase of administration. When samples are drawn too early after the start of administration, the assumption of instantaneous mixing is violated; the concentrations may be higher than expected, and the central volume of distribution may be underestimated. In our data, the first sample was not drawn before 2 min after the start of administration (even in the case of bolus administration), whereas Schnider et al. 3and Dyck and Shafer 2measured propofol 1 min or even 30 s after the start of infusion. These differences in sampling may explain partially the different estimates of central volume of distribution. For the Oostwouder et al. 4data, the early sampling is unclear because this study has not been published in a peer-reviewed journal but only as an abstract with limited information.
Therefore, we have one parameter set from one study and other parameters from other studies, and it is only a hypothesis that one parameter set is superior to the other. One can turn the arguments of Vuyk et al. the other way around. If we calculate a TCI infusion scheme with the Schnider et al. 3data and predict the resulting concentrations with our parameter set, we find an underdosing at the beginning and an accumulation (overdosing) toward the end of anesthesia (fig. 1). If we calculate the ratio (measured concentration)/(predicted concentration) for the elderly individuals of our data set with the parameters of Schnider et al. , 3the concentrations are underestimated (fig. 2). Again, this is not proof that our results are right and the others are wrong, but the reverse claim of Vuyk et al. has no more evidence.
The more interesting question, however, is related to the consequences for dosing in clinical practice. Vuyk et al. claim that the use of our pharmacokinetic parameters for TCI “even may be harmful” to elderly patients because of an overdosing in the initial infusion phase and the resulting hemodynamic depressive effects of propofol. In many countries (European Union, Australia, New Zealand, South Africa), a commercial TCI system has been available to patients for several years (Diprifusor-TCI; Astra Zeneca Pharmaceuticals, Wilmington, DE). This system, which is approved for patients between 16 and 100 yr, uses the pharmacokinetic data from Gepts et al. 6with a central volume of approximately 17 l for a patient weighing 75 kg, irrespective of the patient’s age. Following the arguments of Vuyk et al. , use of this system in elderly patients would be extremely dangerous because of a fourfold overdosing during induction. However, several million applications have been performed with this TCI system in the past 3 yr, but there were no more hemodynamic problems during induction with this system than with conventional dosing strategies (H. Brasch, M.D., Astra Zeneca, written communication, July 2000).
Finally, we would like to focus on a more general aspect of TCI. It often is pointed out that the targeted concentrations are never achieved exactly because the error of such systems is typically in the range of 25–30%. Therefore, the term “target-controlled” is misleading. In contrast, the first description of a computer-controlled infusion device by Schüttler et al. 7as CATIA (Computer Assisted Titration of Intravenous Anesthesia) might have been more appropriate because the aspect of titration was more emphasized. Even with a TCI system, the anesthetist has to find out the optimal dosing by careful titration, and computer-controlled infusion systems can facilitate this process. An experienced anesthetist would never choose the “one and only” optimal value of a fixed blood concentration for every patient, particularly if the patient is elderly or has polymorbidity.
Vuyk et al. speculate with more or less reasonable arguments in reflection on our data, but valid proof to support their opinions is missing. However, this letter shows that pharmacokinetic modeling of propofol is still of interest and that there is a need for prospective validation, especially in regard to its application to TCI.