1 June 2008
Open Letter To: Tony Absalom; Jan Henning Austnes; Luc Barvais; Frank Engbers; Mats Enlund; JuanCarlos Esquivel; Siv Catherine Hoymark; Harald Ihmsen; Gavin Kenny; Wilfried Mertens; Hugh Pulsford; Johan Raeder; Ann Rigby-Jones; Stéphane Ruton; Thomas Schnider; Stefan Schraag; Jürgen Schüttler; Frédérique Servin; Robert Sneyd; Michel Struys; Nick Sutcliffe; Alastair Thomson; Jaap Vuyk; Wolfgang Weyh; Martin White;Catarina Nunes; Valérie Billard; Tomiei Kazama; Tom de Smet; Steve Shafer.
From: Dr J B Glen.1
Re: Copenhagen TCI Workshop, June 1, 2008. 2
The following topics were discussed:
History of commercial TCI developments
Iain Glen reminded the group that in 1992 when Zeneca decided to develop a commercial Target Controlled Infusion (TCI) system, research systems had been developed in the US (Stanford and Duke), Glasgow, Holland, Belgium and France following the pioneering work of Helmut Schwilden and Jürgen Schüttler in Bonn. There were also many different PK models for propofol being used in these systems. For commercial development there was a need to standardize. The selection of the model published by Marsh et al. (hereafter called the ‘Marsh model’) was based on two studies which compared predictive performance with this model with other favoured models proposed by Dyck and Shafer and Tackley et al. The software developed in Glasgow was selected based on the extensive clinical experience which had already been gained with this implementation and the security provided by a two processor design.
A prospective Zeneca sponsored clinical trial programme with this equipment in more than 400 patients provided information on target blood propofol settings for inclusion in Diprivan drug labelling. Predictive performance of the Marsh model was assessed in two studies and indicated a degree of positive bias which was considered clinically acceptable (Swinhoe et al. – 46 patients, MDPE 16.2%: Barvais et al. – 21 patients, MDPE 21.2%). A regulatory process was established whereby both drug and device authorities were involved.
Many thousands of Diprifusor TCI systems are now used in most countries of the world (USA and Canada are still exceptions) and the system has proved to be clinically effective and robust. Shortly after the first systems were introduced a ke0 of 0.26/min was added to allow the display of predicted effect-site concentration. This figure was based on the results of a preliminary analysis of a study by Martin White et al. presented as an abstract at the 11th World Congress of Anaesthesiologists (Sydney, 1996). Effect-site control was not considered initially because the regulatory process for this novel drug/device combination was already considered complex and as time progressed there was uncertainty as to which ke0 to use with the Marsh model.
The commercial ‘buy-in’ to the Diprifusor TCI development was obtained only on condition that this would be restricted to use with electronically tagged Zeneca prefilled Diprivan syringes. The consequence of this strategy was that once generic propofol became available there was a demand for ‘open TCI’ systems that would deliver generic propofol. This led to the development of the Fresenius Kabi ‘Base Primea’ system and the Cardinal Health ‘Asena PK’ pump. In the meantime, PK modelling had progressed and the model published by Schnider et al. was considered attractive because it required covariates for age, weight and lean body mass and because they had developed a ke0 for their model in an integrated PK/PD study. Thus, the situation arose that alternative models for propofol became available. These devices were approved under device legislation without any reference to the pharmaceutical company responsible for drug labelling.
As more experience has been gained with the different models and implementations it is clear that significant differences in drug delivery occur. The confusion arising is damaging the concept of TCI and is driving the need for the Open TCI Initiative to reach a consensus on a single preferred model for propofol.
The implications of differences between currently used models
The Marsh model, at any given target setting delivers the same amount of drug on a mg/kg basis independent of a patient’s age or weight. Elderly patients are easily managed by setting a lower target and obese patients by setting a lower target or inputting a lower patient weight. However, with advances in PK modelling it would be desirable for dosage to be individualised with patient covariates. It is also acknowledged that there are clinical advantages in effect-site control and this is not currently available in Diprifusor systems.
Martin White presented a summary of his recent study in which he has developed age and gender covariates for the central volume of distribution (V1) and clearance while fixing the other parameters of the Marsh model (hereafter referred to as the ‘White model’). Bias was reduced from 12.6 % with the Marsh model to 3.2% with the White model.
Differences in the initial loading dose delivered by different models (when used in plasma control mode) are related directly to differences in V1. The White model has a smaller V1 than the Marsh model and thus delivers a smaller loading dose. The model published by Schnider et al. delivers a very much smaller loading dose in plasma control, explaining the preference to use this model in effect site control where the loading dose approaches that calculated by the Marsh model.
Frank Engbers illustrated some of the differences resulting from the different implementations of effect-site control with the model published by Schnider et al. in Fresenius Kabi and Cardinal Health systems. As these systems are used at the extremes of patient weight the limitations of the James equations for lean body mass become evident with values decreasing above a certain body weight, dependent on height and gender. Both Fresenius Kabi and Cardinal Health have now implemented software changes which warn the user that this model should not be used in male patients with a body mass index (BMI) > 42 kg/m2 or in female patients with BMI > 35 kg/m2.
Helmut Schwilden made the point that some form of scaling is probably required for obese patients but lean body mass may not be the best way. Other equations for lean body mass have been described and Steve Shafer reports that the formula described by Janmahastian et al. appears to perform in a more logical way. Also, Hume described the linear regression of lean body mass on weight and height for male and female patients. In future studies it might also be useful to look at BMI as a possible covariate.
The results of a recent comparision of the predictive performance of Diprifusor (the Marsh model), with the White model, and with the models published by Schüttler et al. and Schnider et al. were presented. Measured arterial propofol concentrations (approx. 30 per patient) were available from 9 patients used as normal controls in a previous study by Servin et al. where a standardised infusion scheme had been used. Values for MDPE and MDAPE calculated for the whole period of the study demonstrated that all four models met the generally accepted criteria for ‘clinical acceptability’. However, when performance at different phases was examined some differences between the models were seen. In the early phase the model published by Schnider et al. tended to overpredict the measured concentrations and to underpredict in the ‘washout’ phase at the end of infusion. The White model showed the smallest bias in the middle ‘maintenance’ phase with predicted concentrations close to those of the model published by Schnider et al. at this stage. Thus a low value for MDPE over all phases may mask opposing trends at certain phases.
Most studies with Diprifusor TCI demonstrate a positive bias of 15-20%. This work confirmed Martin White’s result that the covariate model would be likely to produce a smaller bias than the current Diprifusor system.
It was suggested that in any new analysis it may be necessary to chose an objective function which is very good initially. It may not be possible with a conventional 3-compartment model to improve accuracy at all phases of administration.
Implementation of a new model
There was widespread agreement on the need for a single model for propofol and strong support for the Open TCI initiative. Thomas Schnider summarised progress to date and encouraged the continued submission of as much data as possible.
There is an expectation that any new PK model will be similar to existing models and existing systems could either be upgraded or replaced over time. Pump companies will require some evidence of agreement on the validity of a new model. There is some concern that the PD component will pose a greater challenge.
Methodology for ke0 determination
Dr J Mourisse presented information[16,17] at the EuroSIVA meeting indicating that effects measured at cortical and subcortical structures have different ke0‘s. It is also important to interpret ke0 in the light of anaesthetic conditions, in particular alveolar ventilation and arterial PaCO2 as marked differences in sevoflurane ke0 can be related to changes in ventilation.
The concept of time to peak effect-site concentration (tpeak)[18-20]being a model independent parameter is being challenged by recent work indicating an influence of injection time. With effect-site TCI, the initial loading dose is generally given over 30-40s depending on the target and the weight of the patient and results obtained with this injection time are likely to differ from those following a 5-10s bolus.
For drugs with rapid onset, model misspecification may result in a poor estimation of the brain exposure such that different ke0s can be expected with different study designs. However, the question remains as to which ke0 value for a given model works best for effect-site TCI with that model. Another approach suggested by Charles Minto3 is that one should calculate the value of ke0 that works best when the pump is running at its maximum rate, i.e. 1200 ml/h.
No firm views were expressed as to whether it is best to use a fixed ke0 with a given model as is used with the model published by Schnider et al. in the Fresenius Kabi pump (Base Primea) or to use a fixed time to peak effect and adjust ke0 accordingly as is done in the Cardinal Health pump (Asena PK).
Methodology to compare the clinical effects of different ke0 values for the same model and different effect-site TCI systems
A view was expressed that for the Marsh model, a ke0 value of 0.2/min as published by White et al. is too slow and a ke0 of 1.2/min as published by Struys et al. is too fast. Struys et al. showed that the maximum BIS effect occurred close to the time predicted by the faster ke0. A surprising observation in this study was that the large loading dose given with the slower value did not lead to a significantly shorter induction time.
Barakat et al.[24,25] found that the time course of the onset of sedation as measured by BIS or the OAAS score followed more closely the effect-site concentrations predicted by the Marsh model with a ke0 of 0.26/min than the concentration predicted by the model published by Schnider et al. with ke0 of 0.456/min.
In another recent study, Alastair Thomson and colleagues in Edinburgh used the Marsh model with a range of ke0 values to achieve level 3 OAAS sedation. The effect site target was increased incrementally and when the desired level of sedation was reached the target was fixed at the calculated effect-site concentration indicated at that time. With a ke0 of 1.2/min, deepening sedation was seen over the ensuing 15 min whereas with ke0 of 0.2/min sedation lightened. There was marked interindividual variability in ke0 with values within the range of 0.5/min to 0.8/min appearing suitable for most patients. A population ke0 value of 0.6/min was calculated using probit analysis based on the number of patients showing deepening sedation. This approach appears to take account of interindividual PK and PD variability and could be applied in other situations, e.g., to maintain a steady effect on BIS during anaesthesia.
Doufas et al. showed that when the correct rate of plasma-effect site equilibration was determined for each individual, the effect site concentrations associated with each clinical measure were not affected by the rate of increase of effect site propofol concentration. They published a population ke0 value of 0.17/min. When compared with the ke0 value of 0.456/min obtained by Schnider et al. in their integrated PK/PD study, there is a marked difference in the initial loading dose (e.g., in a 70kg, 170cm, 50yr male patient based on a target of 4 mcg/ml; 1.8 mg/kg with ke0 = 0.17/min vs. 0.9 mg/kg with ke0 = 0.456/min.). Interestingly the Marsh model with a ke0 = 0.6/min provides a loading dose of 1.7 mg/kg in this same patient, similar to that delivered with the slower value for the Schnider model.
For the comparison of different ke0s for the same model, Helmut Schwilden suggested studies where the benefits of effect-site TCI should be seen, e.g., faster onset. He also suggested that if a feed back control system was linked effect to effect-site concentration, the best ke0 would be the one which led to the smallest degree of oscillation when anaesthetic depth was changed.
A study is planned in Edinburgh to compare anaesthesia induction times with effect-site TCI with the ke0 of 0.6/min obtained by Thomson et al. for the Marsh model with the faster ke0 value of 1.2/min published by Struys et al. (based on tpeak = 1.6 min) and the integrated PK/PD model published by Schnider et al.
Few conclusions were reached but hopefully the understanding of some of the differences between models and methodologies was facilitated. There is widespread agreement on the need for a single model for propofol and support for the Open TCI Initiative.
Dr J B Glen
Glen Pharma Ltd
- Marsh B, White M, Morton N, Kenny G: Pharmacokinetic model driven infusion of propofol in children. Br J Anaesth 1991; 67:41–8 PMID: 1859758
- Dyck B, Shafer S: Effects of age on propofol pharmacokinetics. Seminars in Anesthesia 1992; 11:2–4
- Tackley R, Lewis G, Prys-Roberts C, Boaden R, Dixon J, Harvey J: Computer controlled infusion of propofol. Br J Anaesth 1989; 62:46–53 PMID: 2783854
- Swinhoe C, Peacock J, Glen J, Reilly C: Evaluation of the predictive performance of a “Diprifusor” TCI system. Anaesthesia 1998; 53 Suppl 1:61–7 PMID: 9640119
- Barvais L, Rausin I, Glen J, Hunter S, D’Hulster D, Cantraine F, Hollander A d’: Administration of propofol by target-controlled infusion in patients undergoing coronary artery surgery. J Cardiothorac Vasc Anesth 1996; 10:877–83 PMID: 8969394
- White M, Engbers F, Schenkels M, Burm A, Bovill J: The pharmacodynamics of propofol determined by auditory evoked potentials. 11th World Congress of Anaesthesiologists 1996:A608
- Schnider T, Minto C, Gambus P, Andresen C, Goodale D, Shafer S, Youngs E: The influence of method of administration and covariates on the pharmacokinetics of propofol in adult volunteers. Anesthesiology 1998; 88:1170–82 PMID: 9605675
- Schnider T, Minto C, Shafer S, Gambus P, Andresen C, Goodale D, Youngs E: The influence of age on propofol pharmacodynamics. Anesthesiology 1999; 90:1502–16 PMID: 10360845
- White M, Kenny G, Schraag S: Use of target controlled infusion to derive age and gender covariates for propofol clearance. Clin Pharmacokinet 2008; 47:119–27 PMID: 18193918
- James WPT: Research on obesity. London. Her Majesty’s Stationary Office 1976
- Janmahasatian S, Duffull S, Ash S, Ward L, Byrne N, Green B: Quantification of lean bodyweight. Clin Pharmacokinet 2005; 44:1051–65 PMID: 16176118
- Hume R: Prediction of lean body mass from height and weight. J Clin Pathol 1966; 19:389–91 PMID: 5929341
- Schüttler J, Ihmsen H: Population pharmacokinetics of propofol: a multicenter study. Anesthesiology 2000; 92:727–38 PMID: 10719952
- Glen J, Servin F: Evaluation of the predictive performance of four pharmacokinetic models for propofol. Br J Anaesth 2009; 102:626–32 PMID: 19297371
- Servin F, Cockshott I, Farinotti R, Haberer J, Winckler C, Desmonts J: Pharmacokinetics of propofol infusions in patients with cirrhosis. Br J Anaesth 1990; 65:177–83 PMID: 2223333
- Mourisse J, Lerou J, Struys M, Zwarts M, Booij L: Multi-level approach to anaesthetic effects produced by sevoflurane or propofol in humans: 1. BIS and blink reflex. Br J Anaesth 2007; 98:737–45 PMID: 17519261
- Mourisse J, Lerou J, Struys M, Zwarts M, Booij L: Multi-level approach to anaesthetic effects produced by sevoflurane or propofol in humans: 2. BIS and tetanic stimulus-induced withdrawal reflex. Br J Anaesth 2007; 98:746–55 PMID: 17519262
- Minto C, Schnider T, Gregg K, Henthorn T, Shafer S: Using the time of maximum effect site concentration to combine pharmacokinetics and pharmacodynamics. Anesthesiology 2003; 99:324–33 PMID: 12883405
- Nigrovic V, Amann A: Time to maximum effect site concentration. Anesthesiology 2004; 100:1319; author reply 1320-1 PMID: 15114235
- van M, Nikkelen E, Good M: Comments on using the time of maximum effect site concentration to combine pharmacokinetics and pharmacodynamics. Anesthesiology 2004; 100:1320; author reply 1320-1 PMID: 15114236
- Struys M, Coppens M, De N, Mortier E, Doufas A, Van B, Shafer S: Influence of administration rate on propofol plasma-effect site equilibration. Anesthesiology 2007; 107:386–96 PMID: 17721240
- White M, Schenkels M, Engbers F, Vletter A, Burm A, Bovill J, Kenny G: Effect-site modelling of propofol using auditory evoked potentials. Br J Anaesth 1999; 82:333–9 PMID: 10434811
- Struys M, De S, Depoorter B, Versichelen L, Mortier E, Dumortier F, Shafer S, Rolly G: Comparison of plasma compartment versus two methods for effect compartment–controlled target-controlled infusion for propofol. Anesthesiology 2000; 92:399–406 PMID: 10691226
- Barakat A, Sutcliffe N, Schwab M: Effect site concentration during propofol TCI sedation: a comparison of sedation score with two pharmacokinetic models. Anaesthesia 2007; 62:661–6 PMID: 17567340
- Schnider T, Minto C: Pharmacokinetic models of propofol for TCI. Anaesthesia 2008; 63:206; author reply 206-7 PMID: 18211459
- Thomson AJ, Nimmo AF, Glen JB: Determination of the optimum ke0 value for use with the Marsh PK model for propofol: 9AP5-7. European Journal of Anesthesiology (Supplement 39) 2007; 24:119
- Doufas A, Bakhshandeh M, Bjorksten A, Shafer S, Sessler D: Induction speed is not a determinant of propofol pharmacodynamics. Anesthesiology 2004; 101:1112–21 PMID: 15505446
- These notes include some post meeting thoughts and some comments from Steve Shafer. ↩
- Following the initial meeting in Cape Town this workshop was organized by Iain Glen to provide a possible ‘back-up’ strategy to select a preferred propofol model from those currently available. However, prior to this meeting, many data sets were contributed to the Open TCI Initiative and there was more optimism that the Open TCI Initiative would be successful in meeting its goals. The objective of the “TCI Workshop” thus changed to provide a forum for those invited by Iain Glen to discuss aspects of TCI that are leading to current debate, as an adjunct to and under the same ground rules as the Open TCI Initiative. It was unfortunate that several of the registered members of the Open TCI Initiative were completely unaware of this meeting. However, it is encouraging to see that many of those who attended the “Copenhagen TCI Workshop” have now registered with the Open TCI Initiative. ↩
- Suggestion made during a conversation with Steve Shafer on a boat trip at the First World Congress of TIVA/TCI in Venice, Italy, September 2007. Unpublished! ↩