Pharmacokinetic/Pharmacodynamic modeling of muscle relaxants
D.M. Fisher. Department of Anesthesia and Pediatrics, University of California, San Francisco, USA

Pharmacokinetic and pharmacodynamic modeling has prospered in anesthesia because of the types of drugs used in our profession, our ability to make meaningful measurements, and our ability to perform high-resolution studies, e.g. the opportunity to sample arterial blood frequently. In the past two decades, much seminal work has been performed using muscle relaxants. In this session, I will review a number of these seminal studies. I make no attempt to provide a comprehensive overview of all issues related to kinetic/dynamic (PK/PD) modeling of relaxants. Instead I review those studies that have impacted most significantly on our understanding of modeling.

Effect compartment modeling
An intuitive approach to modeling effect of a drug that works directly is to assume that the plasma concentration of the drug relates to effect. If drug concentration changes quite slowly (e.g. if the drug is administered orally and measurements are made only at steady state), this assumption is reasonable. However, most anesthetic drugs and most muscle relaxants are not used in this manner. Hence, using the plasma concentration to represent effect site concentration is presumably flawed. Although the introduction of this concept to modeling is credited to Segre, widespread application of this concept in all of pharmacology (i.e. not just in anesthesia) resulted from seminal manuscripts by Hull et al.[1]and by Sheiner et al.[2] Of note, Sheiner et al. administered d-tubocurarine by brief infusion so as to obtain information about the largest fraction of the Cp/effect hysteresis curve.

Semiparametric modeling
Most PK/PD modeling assumes that the plasma concentration vs. time curve can be described as the sum of exponentials. In many situations this is appropriate. However, if the time shortly after drug administration is critical (as it typically is for anesthetic drugs), then the assumption is flawed. Studies by Henthorn et al.[3] demonstrate that arterial plasma concentration of drugs administered intravenously increases during the first 30 sec, then oscillates before decreasing monotonically; this contrasts to the continuous monotonic decrease mandated by traditional exponential models. If Cp is not described appropriately by the sum of exponentials, then PD based on exponential models for Cp is flawed. Fortunately, techniques exist to address this issue. If arterial plasma samples are obtained at the appropriate intervals after drug administration (e.g. at critical times during the first min), the Cp vs. time curve can be described by a series of splines (smoothed curves) or lines connecting the measured plasma concentration values. The rate constant keo is then used to convolve these plasma concentrations (in a manner similar to that with the exponential models) to estimate the concentration in the effect compartment. If a drug is administered by bolus and if effect occurs sometime before Cp decreases monotonically, then this approach to necessary to obtain meaningful values for keo.

Modeling without Cp data
The traditional approach to modeling PK/PD is to measure both plasma concentrations and effect, then to model the dose/Cp relationship (pharmacokinetics) and the Cp/effect relationship (equilibration, pharmacodynamics). However, this is costly because it requires measurement of Cp. Verotta[4] proposed that the sequential PK/PD model could be modified to estimate the steady state infusion rate that produced a certain effect (for example, the infusion rate depressing twitch tension 50%, IR50), a parameter of more clinical utility than either plasma clearance or the steady state concentration depressing twitch tension 50%. Data from Donati et al.[5,6] for laryngeal muscle function and adductor pollicis twitch depression provided Bragg et al.[7] the opportunity to evaluate whether this could be applied to real data. This study provided estimates of both the relative sensitivities of the two muscle groups and the relative rates of equilibration and offered insight into the time course of paralysis at the two muscle groups. However, the analysis suggested one aberrant finding, that IR50 varied with the initial bolus dose. To verify or refute this finding, we[8] repeated Donati's studies with measurements of arterial plasma concentrations of the muscle relaxant. To our surprise, we learned C50 (and IR50) varied with dose, a finding for which no explanation is presently available. The ability to model effect without plasma concentration represents an extreme example of the sparse plasma sampling approach available with population techniques.

The role of metabolites
Segredo et al.[9] study in critically ill patients demonstrated that vecuronium's metabolite cumulated in patients with renal failure and probably contributed to the prolonged paralysis seen with chronic administration of vecuronium. However, verification of this hypothesis depended on estimation of the potency of this metabolite compared to the parent compound. An elegant study by Caldwell et al.[10] determined this potency ratio by administering the parent compound and the metabolite to volunteers using a crossover design. A less powerful design, administering parent compound to one group and metabolite to another group, was used recently by Schiere et al.[11] to estimate the potency of the metabolite of rapacuronium.
Modeling of antagonists: All the pharmacodynamic models discussed previously assume that the concentration of the drug (either in plasma or at the effect site) produces the effect directly. For muscle relaxants, in which the mechanism of effect is well known, this assumption is reasonable and can be modeled as:
Dose => Cp <=> Ce <=> Effect                                 (1)

However, for neostigmine, the relationship is more complicated. Administration of neostigmine produces a plasma (and effect site) concentration which "poisons" cholinesterase. In turn, the decreased activity of cholinesterase increases the concentration of acetylcholine at the neuromuscular junction, thereby producing an effect. This can be summarized as:

Dose => Cp <=> Ce => Cholinesterase Inhibition => Acetylcholine (2)
Muscle Relaxant => Effect                      (3)

Thus, there is no direct effect of the antagonist; instead, the time course of recovery is mediated by the regeneration of cholinesterase rather than the rate at which plasma neostigmine concentration's decrease. This is know as an indirect model. These models have been invoked only rarely in anesthesia[12] (but frequently in other areas, such as coagulation); however, they are necessary to explain certain types of effects.

In summary, kinetic/dynamic modeling has been used extensively in the area of muscle relaxants. The availability of well-defined and easy-to-obtain effect measures, the ability to sample arterial blood, and the knowledge of the "model" have permitted significant advances in this area.

References
1. Hull CJ, Van Beem HB, McLeod K, Sibbald A, Watson MJ: A pharmacodynamic model for pancuronium. Br J Anaesth 1978;50:1113-23
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;25:358-71
3. Henthorn TK, Avram MJ, Krejcie TC, Shanks CA, Asada A, Kaczynski DA: Minimal compartmental model of circulatory mixing of indocyanine green. Am J Physiol 1992;262:H903-10
4. Verotta D: An inequality-constrained least-squares deconvolution method. J Pharmacokinet Biopharm 1989;17:269-89
5. Donati F, Meistelman C, Plaud B: Vecuronium neuromuscular blockade at the diaphragm, the orbicularis oculi, and adductor pollicis muscles. Anesthesiology 1990;73:870-5
6. Donati F, Meistelman C, Plaud B: Vecuronium neuromuscular blockade at the adductor muscles of the larynx and adductor pollicis. Anesthesiology 1991;74:833-7
7. Bragg P, Fisher DM, Shi J, Donati F, Meistelman C, Lau M, Sheiner LB: Comparison of twitch depression of the adductor pollicis and the respiratory muscles. Pharmacodynamic modeling without plasma concentrations. Anesthesiology 1994;80:310-9
8. Fisher DM, Wright PM: Are plasma concentration values necessary for pharmacodynamic modeling of muscle relaxants? Anesthesiology 1997;86:567-75
9. Segredo V, Caldwell JE, Matthay MA, Sharma ML, Gruenke LD, Miller RD: Persistent paralysis in critically ill patients after long-term administration of vecuronium. N Engl J Med 1992;327:524-8
10. Caldwell JE, Szenohradszky J, Segredo V, Wright PM, McLoughlin C, Sharma ML, Gruenke LD, Fisher DM, Miller RD: The pharmacodynamics and pharmacokinetics of the metabolite 3-desacetylvecuronium (ORG 7268) and its parent compound, vecuronium, in human volunteers. J Pharmacol Exp Ther 1994;270:1216-22
11. Schiere S, Proost JH, Wierda JMKH: Pharmacokinetics and pharmacokinetic/ pharmacodynamic (PK/PD) relationship of ORG 9488, the 3-desacetyl metabolite of ORG 9487 (abstract). Anesthesiology 1997;87:A377
12. Verotta D, Kitts J, Rodriguez R, Coldwell J, Miller RD, Sheiner LB: Reversal of neuromuscular blockade in humans by neostigmine and edrophonium: a mathematical model. J Pharmacokinet Biopharm 1991;19:713-29