2004 Annual Scientific Meeting - CLICK FOR PROGRAMME

PK/PD - Why does it matter?

F.Engbers, Leiden, The Netherlands. Email: fengbers@wxs.nl 

Application of a pharmacokinetic model to control an infusion pump has for a long time only been accessible for research. The introduction of Target Controlled Infusion for propofol has made this technique available for the clinician and many thousands patients have been successfully anaesthetised using a computerised infusion pump that calculated the required infusion rate based on a pharmacokinetic model: the Diprifusor.

The Diprifusor holds a single three-compartment pharmacokinetic model that has only a for weight adjusted central volume No other patient characteristics influences the model and in the first release even no information on the effect compartment was available. The change of the central volume according to weight originates from the habit to express the central volume in ml per kg units.  The model used is derived from samples of several hundred patients and what is important: this TCI system has been prospectively tested.

Recently new Target Controlled Infusion devices have come to the market. These new pumps differ from the Diprifusor in a number of ways:

The possibility to use TCI for more drugs is without doubt a big advantage. On the other hand the mixture of different pharmacokinetic data sets and different keo’s for propofol is at least confusing for the user.

For the determination of the performance of a TCI system with a specific pharmacokinetic model measured concentrations have to be compared with predicted concentrations.

Difference between measured and predicted is expressed as the

Prediction Error:

PE=(Measured – Predicted)/Predicted *100

With this prediction error the following can be calculated¹

·        Bias or Median Prediction Error        
MDPE= Median{PEj,j=1,…,N}

·        Precision or Absolute Median Prediction Error
MDAPE=Median{|PEj|,j=1,…N}

·        Divergence : time related change in the absolute prediction error
=linear regression of |PE|

·        Wobble : median of the absolute difference between the prediction error and the bias
=Median{|PEj-MDPE|,j=1,…,N}

When the clinical usefulness of a specific model for TCI application is judged then some aspects of these performance values have to be considered.

In the Diprifusor the calculated concentration in the effect compartment was only used for information on the delay time between achieved blood concentration and the observed effect.

With the introduction of targeting the effect site instead of the blood compartment the blood brain equilibration constants (keo) becomes part of the dosing algorithm which puts an even greater emphasis on the correctness of this constant

When looking at literature, a wide variety of Keo values for propofol have been published ^{3 ,4 ,5,6}. Corresponding blood-brain equilibration half-lives(Thalfkeo=0.693 / keo) vary from short to intermediate long. The Keo ‘connects’ blood concentrations to effect. When instead of measured blood concentrations predicted blood concentrations are used then the Keo becomes part of the pharmacokinetic model with which these predictions are made. Three compartment models are static, they assume immediate mixing of the drug in the central compartment, which is of course not true. Therefore inaccuracies of the model that will be prominent with large concentration changes will become part of the keo when such a model is used to resolve the keo. Therefore it is not correct to use a keo that belongs to one pharmacokinetic set to calculate the effect compartment concentration in another pharmacokinetic model for the same drug. (When no keo is available for a specific model then theoretically the keo that is fitted with a non parametric approach is the best to use)   Using different pharmacokinetic data sets will lead to different keo’s but this only explains partly why the variation in reported keo’s for propofol is so big.

The traditional way to measure the keo is by constructing an hysteresis loop between the blood concentration and the measured effect². The keo measured using the hysteresis loop is dependent on both an increase and a decrease of the blood concentration. Recently another approach has been suggested. After a bolus the time required to reach the peak of the effect is measured. From this time to peak effect the keo can be calculated. Using another pharmacokinetic model to predict the concentration time course will obviously result in a different keo. Using the time to peak effect method the keo from one pharmacokinetic dataset can be easily transferred to another dataset⁷. But this is in theory. Time to peak effect uses only one measured effect point (the maximum) from the effect-time curve. In contrast to the hysteresis method the shape of the effect curve does not influence the keo.

But maybe more important: the time to peak effect method is only usable for bolus studies. Just recently some studies have shown that for propofol the keo appears to be infusion rate dependent^8,9  If this is the case than the time to peak effect will lead to an overestimation of the keo that would not be usable for infusion  as in TCI. On the other hand the keo used in the Diprifusor that clinically seems to function reasonable well may then be too small to be used for effect site control⁵.

But: using an ‘incorrect’ too big keo (short Thalfkeo) for effect site control may in certain aspects clinically be more appropriate then using the ‘correct’ median keo from the population to which the patient belongs. This is because a bigger keo will reduce the loading dose when the target is increased and also will reduce the time the infusion stops when the target is decreased. On the other hand useful information provided by the effect site is now lost¹⁰

Effect TCI will without any doubt be the next development in TCI. It is however a pity that with so many unsolved questions, marketing pressure seems to prevail above proper  prospective testing.

1.                Varvel JR, Donoho DL, Shafer SL: Measuring the predictive performance of computer controlled infusion pumps. J Pharmacokinet Biopharm 1992; 20:63-94 

2.                Sheiner LB, Stanski DR, Vozeh S, Miller RD, Ham J. Simultaneous modeling of pharmacokinetics and pharmacodynamics: application to d-tubocurarine. Clin Pharmacol Ther 1979;25:358-371.

3.                Schnider T, Minto C, Shafer S et al: The influence of Age on Propofol Pharmacodynamics. Anaesthesiology 90(6):1502-1516,1999

4.                White M, Schenkels MJ, Engbers FH, et al.
Effect-site modelling of propofol using auditory evoked potentials.
Br J Anaesth (England), Mar 1999, 82(3) p333-910.

5.                Struys M, De Smet T, Depoorter B et al: Comparison of Plasma Compartment versus Two Methods for Effect Compartment-controlled Target-controlled Infusion for Propofol. Anesthesiology 92(2):399-406, 2000

6.                Predictability of processed electroencephalography effects on the basis of pharmacokinetic-pharmacodynamic modeling during repeated propofol infusions in patients with extradural analgesia.
Kuizenga K, Proost JH, Wierda JM, Kalkman CJ.
Anesthesiology. 2001 Sep;95(3):607-15

7.                Using the time of maximum effect site concentration to combine pharmacokinetics and pharmacodynamics.
Minto CF, Schnider TW, Gregg KM, Henthorn TK, Shafer SL.
Anesthesiology. 2003 Aug;99(2):324-33

8.                The Influence Of Propofol Administration Rate On The Ke0 Value. Preliminary Results. Michel MRF Struys, MD,PhD ; Nicolaas De Neve, MD ; Tom De Smet, MSc ; Dorine Dyzers, CN; Eric P Mortier, MD, DSc and Steven L. Shafer, MD. Abstract at the Eurosiva meeting Lisbon 2004(http://wwweurosiva.org)

9.                Factors of influence on Pk-Pd of IV anaesthetics.
Tomiei Kazama, MD
Department of Anesthesiology, National Defense Medical College
3-2 Namiki, Tokorozawa, Saitama, Japan 359-8513. Abstract at the Eurosiva meeting Lisbon 2004 (http://www.eurosiva.org)

10.            Calculated effect-site propofol concentration at loss, and regaining, of consciousness: as predicted by two different pharmacokinetic models.
J. Prinzlin*, N. Sutcliffe*, F. Engbers**
* Golden Jubilee Hospital Clydebank Glasgow
** Leiden University Medical Center