EUROSIVA, Nice 2002                         What is new regarding NMBA administration

 

Bayesian forecasting and closed loop control of NMBA

Dr Valerie Billard[*]

 

 

 

Why should we adjust muscle relaxants?

By providing skeletal muscle relaxation, neuromuscular blocking agents (NMBA) are useful in anaesthesia to facilitate intubation, and to allow safe abdominal and thoracic surgery.

They may also be recommended in ICU to decrease oxygen consumption and improve mechanical ventilation efficacy.

 

The "ideal agent" , or at least the ideal dose regimen of the existing agents should combine fast onset to increase quickly the relaxation if necessary, and rapid recovery because neuromuscular function should be back to normal at the end of anaesthesia (1).

 

The level of relaxation required is quite well defined, and could be assessed by neuromuscular monitoring : usually no response visible to TOF for intubation and no more than one response for surgery. Quantitative assessment are available based on the response force following a standardised stimulation, EMG or acceleration.

An insufficient blockade results in poor intubating or operating conditions, and may induce morbidity. An excessive blockade has few consequence as far as the patient is unconscious and ventilated but may increase postoperative morbidity, and induce post traumatic psychological disorders. It also needlessly increases the cost through drug consumption and O.R. occupancy. Both these risks make the need for a good control of the relaxation essential in clinical practice.

 

The intubation dose is usually between 2 and 3 times the dose necessary to paralyse 95% of the patients (ED95) in order to shorter the onset with keeping reasonable recovery times.

It is chosen "a priori" and not adjusted except to body weight. However, monitoring already displays quite large differences in the onset between individuals.

 

The doses required to maintain the expected blockade are more tricky to determine, and the infusion rates can vary in a very wide range for several reasons :

1.Some drugs are cumulative and compensation of distribution needs repeated adjustments over time (1)

2.Interindividual variability may be due to physiological covariates as age, hepatic or renal function, or genetically determined enzymatic equipment (cholinesterases)

3.Intraindividual variability has been described, specially in ICU (2).

4.Pharmacodynamic variability is also wide between patients as well as over time in a same patient, due to drug interactions (volatile agents) or temperature changes.

 

Sophisticated pharmacokinetic (PK) – pharmacodynamic modelling (PD) using population approach can take account of the first factor and partly of the second.

The other factors could only be detected by measuring the blockade and adjusting the dose to achieve the desired blockade. Several ways to do so have been described.

 

Two mains methods may be distinguished :

-         direct (and repeated) adjustments to the measured effect

-         use of PKPD models adjusted to individuals

 

 

Direct adjustment to repeated measures

 

Adjusting the infusion rate to the monitoring could be done manually… and totally empirically as we usually do in clinical practice.

 

It could also be automated in closed loop controllers. In this approach, the body is considered as a black box, the dose of NMBA being the input and the level of blockade the output. This situation looks like many industrial or household devices features and could be solved the same way. A controller calculates the difference  between the measured output and the desired output (let's call it the error "e"), and correct the input according to a preset algorithm to minimize this difference.

 

Most of times, the algorithm to be efficient needed to consider not only e but also how fast it changed (derivative) and what was its overall time course (integral) (3).

Thus, the infusion rate algorithm looks like :

 

v(t) = weight.[ kp.(e) + ki.ň edt+kd . de/dt]

 

Numerous studies during the past 15 years used this kind of algorithm to administer non depolarising NMBA in anaesthesia (4-9) or in ICU (10), most of them using EMG as the measure of effect. All obtained a good control of the blockade (mean e < 10%), despite electrical disturbances, changes in temperature or blood loss.

 

 

Model driven adjusted anesthesia

 

In this approach, initial bolus doses and infusion rates are calculated based on a PKPD model (usually 2 compartment PK model and Emax PD model). Then, some parameters of the PKPD model are adjusted according to the difference between the measured blockade and the blockade predicted by the model.

 

These methods may be included among the bayesian techniques, derived from the Bayes theorem about conditional probabilities (11). In the Bayes theorem, the probability of predicting an event A (which is dependant on an event B) will increase if we know the status of event B.

 

In pharmacology, the probability of being in the therapeutic window after a dose B will increase if the effect of the previous dose A is known, even if it was not by itself in the therapeutic range. These methods have been used to adjust the dose of drugs having a low therapeutic window to a measured concentration in the same patient as for aminosides, vancomycine, theophylline, methotrexate, or CPT11, in adults or children (12). In anesthesia it has been first described for alfentanil (13).

 

For muscle relaxants, a quantitative measure of effect, available in real time made the adjustment to the concentration useless : the PK or the PD model can be adjusted to a patient by taking account of a few measured values of neuromuscular blockade.

Bayesian adjustment may be the done manually as in Stanpump software (14) or can be incorporated in closed loop systems (15-19). The results were so stable and reproducible that some groups used closed loop controller in clinical research to demonstrate drug interactions. They fixed the desired level of blockade (usually 90%), and adjusted the infusion rate using the closed loop for different end-tidal concentrations of volatile agents, different temperature or to examine the influence of CPB on the NMBA requirements.

 

The bayesian approach may appear as too sophisticated compared with the direct adjustment. However it offers several theoretical advantages :

1)      the number of measures may be lower than in the direct adjustment. Few measures will adjust the model regarding interindividual variability, then the model will stay adjusted to that patient even if the measured effect is lost. Further measures of effect may be useful to correct the model according to intraindividual variability (due to drug interactions, temperature changes,…)

2)      The closed loop system adjusts the model to the individual, and the corresponding parameters can be saved and analysed. Thus, correlation to physiological factors (age, renal or liver function, temperature, CPB(18)) or to pharmacological factors (interactions with volatile agents (17;20;21)) may be easily performed.

 

However, the model based adjustments need the a priori choice of  a PKPD model and their performance may be disappointing if the model chosen is wrong (22) or if active metabolites participate in the effect.

 

 

Application of fuzzy logic to NMBA administration

More recently, control of NM blockade using fuzzy logic approach (23) has also been described for computer controlled of NMBA. In this methods, the quantitative difference between measured and target blockade (e) is transformed ("fuzzified") to a qualitative variable such as big or small, positive or negative (24). Then, the algorithm translates this result ("defuzzifies") in a infusion rate change as for example :

-         "if difference is big and positive : stop the pump"

-         " if the difference is big and negative, give a bolus"

Some of these controllers could even modify automatically their algorithm according to the data they receive (self learning fuzzy logic controllers) (25;26).

Several approach may be combined , by using fuzzy controller with PID or model mased output algorithm (27-29).

Based on the few studies published fuzzy controller seems to have similar performance than both other types of control. However, some signal loss have been described after prolonged functioning. (25)

 

 

Clinical implications

 

To summarize, closed loop administration of NMBA may be done by many methods, all appearing efficient despite some signal loss during prolonged infusion, or difficulties to stabilize the system(30).

However, all the closed-loop techniques described here are part of research programs and can be piloted only by a very narrow group of people in the world. Of course, none of these program is today CE. A routine use of closed loop technique may be developed but some potential clinical benefit should first be demonstrated.

While the clinical benefit of monitoring the neuromuscular blockade is admitted, specially to optimise recovery, there are no data about the clinical benefit of closed loop systems vs. manual ("empirical") adjustment to the monitoring. In intensive care, a first study failed to demonstrate that a closed loop infusion of NMBA improved mechanical ventilation of septic shock (10).

In anesthesia, we may expect closed loop to decrease the working charge, improve surgical condtions, and shorten recovery but this need to be verified.

References

 

1.   Donati F. Cumulation and flexibility with infusions of neuromuscular blocking drugs. Can J Anaesth 2000; 47: 936-42.

2.   Segredo V, Caldwell JE, Wright PM, Sharma ML, Gruenke LD, Miller RD. Do the pharmacokinetics of vecuronium change during prolonged administration in critically ill patients? British Journal of Anaesthesia 1998; 80: 715-9.

3.   Billard V, Mavoungou P. Computer-controlled infusion of neuromuscular blocking agents. In: Vuyk J, Engbers F, Groen-Mulder S, eds. On the study and practice of intravenous anaesthesia. Dordrecht: Kluwer Academic Publishers, 2000: 159-72.

4.   Webster NR, Cohen AT. Closed-loop administration of atracurium. Steady-state neuromuscular blockade during surgery using a computer controlled closed-loop atracurium infusion. Anaesthesia 1987; 42: 1085-91.

5.   McLeod AD, Asbury AJ, Gray WM, Linkens DA. Automatic control of neuromuscular block with atracurium. British Journal of Anaesthesia 1989; 63: 31-5.

6.   O'Hara DA, Derbyshire GJ, Overdyk FJ, Bogen DK, Marshall BE. Closed-loop infusion of atracurium with four different anesthetic techniques. ANESTHESIOLOGY 1991; 74: 258-63.

7.   Assef SJ, Lennon RL, Burke MJ, Behrens TL. A versatile, computer-controlled, closed-loop system for continuous infusion of muscle relaxants. Mayo Clin Proc 1993; 68: 1074-80.

8.   Stinson LW, Murray MJ, Jones KA et al. A computer-controlled, closed-loop infusion system for infusing muscle relaxants: its use during motor-evoked potential monitoring. J Cardiothorac.Vasc.Anesth 1994; 8: 40-4.

9.   O'Hara DA, Hexem JG, Derbyshire GJ et al. The use of a PID controller to model vecuronium pharmacokinetics and pharmacodynamics during liver transplantation. Proportional-integral- derivative. IEEE Trans.Biomed Eng 1997; 44: 610-9.

10.       Freebairn RC, Derrick J, Gomersall CD, Young RJ, Joynt GM. Oxygen delivery, oxygen consumption, and gastric intramucosal pH are not improved by a computer-controlled, closed-loop, vecuronium infusion in severe sepsis and septic shock. Crit Care Med 1997; 25: 72-7.

11.       Jelliffe RW, Schumitzky A, Bayard D et al. Model-based, goal-oriented, individualised drug therapy. Linkage of population modelling, new 'multiple model' dosage design, bayesian feedback and individualised target goals. Clin Pharmacokinet. 1998; 34: 57-77.

12.       Fernandez de Gatta MM, Garcia MJ, Lanao JM, Dominguez-Gil A. Bayesian forecasting in paediatric populations. Clin Pharmacokinet 1996; 31: 325-30.

13.       Maitre PO, Stanski DR. Bayesian forecasting improves the prediction of intraoperative plasma concentrations of alfentanil. ANESTHESIOLOGY 1988; 69: 652-9.

14.       Devys JM, Billard V, Barreau-Pouhaer Let al. Administration des curares pour chirurgie plastique : apports de l'adaptation bayésienne. Ann.Fr.Anesth.Reanim. 15[6], R279. 1996.

15.       Ebeling BJ, Muller W, Tonner P, Olkkola KT, Stoekel H. Adaptative feedback-controlled infusion versus repetitive injections of vecuronium in patients during isoflurane anesthesia. Journal of Clinical Anesthesia 1991; 3: 181-5.

16.       Schwilden H, Olkkola KT. Use of a pharmacokinetic-dynamic model for the automatic feedback control of atracurium. Eur J Clin Pharmacol. 1991; 40: 293-6.

17.       Kansanaho M, Olkkola KT, Wierda JM. Dose-response and concentration-response relation of rocuronium infusion during propofol-nitrous oxide and isoflurane-nitrous oxide anaesthesia. Eur J Anaesthesiol 1997; 14: 488-94.

18.       Kansanaho M, Hynynen M, Olkkola KT. Model-driven closed-loop feedback infusion of atracurium and vecuronium during hypothermic cardiopulmonary bypass. J Cardiothorac Vasc Anesth 1997; 11: 58-61.

19.       Uys PC, Morrell DF, Bradlow HS, Rametti LB. Self-tuning, microprocessor-based closed-loop control of atracurium- induced neuromuscular blockade. British Journal of Anaesthesia 1988; 61: 685-92.

20.       Hemmerling TM, Schuettler J, Schwilden H. Desflurane reduces the effective therapeutic infusion rate (ETI) of cisatracurium more than isoflurane, sevoflurane or propofol. Can J Anaesth 2001; 48: 532-7.

21.       Olkkola KT, Kansanaho M. Quantifying the interaction of vecuronium with enflurane using closed- loop feedback control of vecuronium infusion. Acta Anaesthesiol Scand. 1995; 39: 489-93.

22.       Laurin J, Nekka F, Donati F, Varin F. Assuming peripheral elimination: its impact on the estimation of pharmacokinetic parameters of muscle relaxants. J Pharmacokinet.Biopharm. 1999; 27: 491-512.

23.       Asbury AJ, Tzabar Y. Fuzzy logic: new ways of thinking for anaesthesia. British Journal of Anaesthesia 1995; 75: 1-2.

24.       Mason DG, Linkens DA, Abbod MF, Edwards ND, Reilly CS. Automated delivery of muscle relaxants using fuzzy logic control. IEEE engineering in medicine and biology 1994; 678-86.

25.       Edwards ND, Mason DG, Ross JJ. A portable self-learning fuzzy logic control system for muscle relaxation. Anaesthesia 1998; 53: 136-9.

26.       Ross JJ, Mason DG, Linkens DA, Edwards ND. Self-learning fuzzy logic control of neuromuscular block. British Journal of Anaesthesia 1997; 78: 412-5.

27.       Kern SE, Johnson JO, Westenskow DR. Fuzzy logic for model adaptation of a pharmacokinetic-based closed loop delivery system for pancuronium. Artif.Intell.Med 1997; 11: 9-31.

28.       Mason DG, Linkens DA, Edwards ND, Reilly CS. Development of a portable closed-loop atracurium infusion system: systems methodology and safety issues. Int J Clin Monit Comput 1996; 13: 243-52.

29.       Mason DG, Edwards ND, Linkens DA, Reilly CS. Performances assessment of a fuzzy logic controller for atracurium-induced neuromuscular block. British Journal of Anaesthesia 1996; 1074-80.

30.       Geldner G, Schwarz U, Ruoff M et al. [Development of a new closed-loop system for controlling mivacurium- induced neuromuscular blockade]. Anaesthesist 1999; 48: 157-62.

31.       Wait CM, Goat VA. Atracurium infusion during paediatric craniofacial surgery. Closed loop control of neuromuscular block. Anaesthesia 1989; 44: 567-70.

 


[*] Dr V Billard :             Département d'Anesthésie, Institut Gustave Roussy;

39 rue Camille Desmoulins ; 94805   Villejuif. France

e-mail :             billard@igr.fr