Pharmacokinetic Origin of Carbamazepine-induced Resistance to Vecuronium Neuromuscular Blockade in Anesthetized Patients

The protocol was approved by the Institutional Review Board of The Cleveland Clinic Foundation and all 21 patients undergoing neurosurgery gave written informed consent before entry into the study. Ten epileptic patients receiving carbamazepine in a daily dose of 1.2-1.6 g for at least 1 month and scheduled for a craniotomy for excision of tumor or seizure foci were assigned to the carbamazepine group. In addition, 11 patients not receiving any anticonvulsant and undergoing craniotomy for a variety of neurosurgical procedures were enrolled in the control group. Patients were excluded from the study if they had hepatic, renal, or neuromuscular diseases, or if they were taking drugs known to affect the neuromuscular function or anticonvulsants other than carbamazepine.

Neuromuscular function was monitored using a Myograph 2000 (Biometer International A/S, Odense, Denmark). The force of contraction of the adductor pollicis was recorded after stimulation of the ulnar nerve at the wrist by four supramaximal square-wave impulses of 200-micro second in duration delivered at 2 Hz every 12 s through surface electrodes. The outstretched arm was enveloped in a cotton blanket and plastic bag in an attempt to minimize heat loss. Patients were monitored with an esophageal stethoscope, electrocardiogram, direct arterial pressure tracing, pulse oximetry, and mass spectrometry. The mechanomyogram was applied before the induction of anesthesia, which was achieved with 3-5 mg/kg intravenous thiopental and 1-2 micro gram/kg intravenous sufentanil. On obtaining a satisfactory control recording of the mechanomyogram, neuromuscular block was induced with 0.1 mg/kg intravenous vecuronium and the trachea was intubated after complete ablation of the twitch. Anesthesia was maintained with 60% nitrous oxide in oxygen, isoflurane to end-tidal concentration of 0.2-0.6%, and a continuous infusion of 0.25-0.50 micro gram *symbol* kg sup -1 *symbol* h sup -1 sufentanil. Ventilation was provided to maintain end-expired P^CO2between 25 mmHg and 30 mmHg.

Plasma vecuronium concentrations were determined from arterial blood samples (3.5 ml) collected in heparinized test tubes via a 20-G cannula inserted in the radial artery. A control blood sample was taken before the injection of vecuronium and additional samples were collected at 1, 2, 3, 5, 7, 10, 20, 30, 45, 60, 90, 120, 180, 240, and 300 min during and after the surgical procedure. Pancuronium does not interfere with the vecuronium assay, because its quaternary ammoniums at positions 2 and 16 are not amenable to oxidation at the potentials used in this study. [12]Therefore, after the first of the four twitches (T1) had recovered to 50-90% of its baseline value, 0.06-0.1 mg/kg pancuronium bromide was administered, when necessary, to maintain surgical relaxation.

To avoid vecuronium ex vivo degradation, blood samples were kept on ice, centrifuged at 2,400 rpm for 4 min, and the plasma acidified with 12 micro liter sulphuric acid 2 mol/l per milliliter of plasma. Samples were kept frozen at -14 degrees C until analysis by high-performance liquid chromatography coupled to electrochemical detection as described previously. [12]In brief, vecuronium and its major metabolites were extracted using Bond-Elut C¹solid-phase extraction cartridges. To allow for better sensitivity, 1.5 ml plasma was extracted for the 7.8- and 15.6-ng/ml concentrations and the patient samples collected at 180, 240, and 300 min. Samples were eluted in silanized glass tubes with 2 x 500 micro liter of 0.01 M sodium perchlorate solution in methanol. The eluents were evaporated and the dry residue was dissolved in 100 micro liter mobile phase. The mobile phase consisted of 0.033 M phosphoric acid (60% vol/vol) and acetonitrile (40% vol/vol) and adjusted with ammonium hydroxide to a final pH of 5.55. The solution was degassed under vacuum and pumped through a prepacked Spherisorb CN (5-micro meter particle size) column (HiChrom, Reading, UK) at a flow-rate of 2 ml/min at a temperature of 30 degrees C. A Coulochem 5100A electrochemical detector (ESA, Chelmsford, MA) linked to a 5010 analytic cell set at potentials of 0.4 V for the first screening electrode and at 0.8 V for the second selective electrode allowed us to quantify vecuronium, its metabolites, and Org 7465, the internal standard. This method is specific and shows good linearity for vecuronium concentrations ranging from 7.8 to 4000 ng, an intraassay precision of 10% and an interassay reproducibility of 5.8%.

Plasma vecuronium concentrations were fitted to a two-compartment pharmacokinetic model (y = Ae^alphat + Be sup - beta t, where y represents the predicted plasma concentration, A and B the intercepts with the y-axis, alpha and beta the complex transfer rates associated with the distribution and the elimination phases respectively) using a nonlinear least-squares fitting program (Siphar, Simed, Creteil, France) with an inverse weighting factor of 1/y (predicted) or 1/y²(predicted) as deemed most appropriate. The following pharmacokinetic parameters were derived using the fitted model [13]: AUC (area under concentration [A/alpha + B/beta]), AUMC (A/alpha²+ B/beta²), Cl (Dose/AUC), MRT (AUMC/AUC), T¹/2 alpha (0.693/alpha), T¹/2 beta (0.693/beta), Vdss (Cl x MRT), Vc (Dose/(A + B)), and Vd beta (Cl/beta).

Two- and three-compartment models were applied to the data. A two-compartment model was retained because in most instances the experimental concentrations in both groups were more closely predicted by using a biexponential rather than a triexponential equation. Our choice of model was further confirmed by using the statistical technique referred as Akaike's information criterion. [14]Finally, a three-compartment analysis required that we have at least four points per phase. This was not always feasible, because we only collected a maximum of 18 samples per patient. Consequently, because the same type of analysis should be chosen for all patients, a two-compartment analysis was selected.

In a model-dependent pharmacokinetic analysis, the accuracy of the clearance and Vdss values will be greatly influenced by the appropriate description of the terminal phase of the concentration-time curve. Consequently, to ensure robustness of the results, these two parameters as well as the total AUC were also calculated by performing a model-independent pharmacokinetic analysis.

In the pharmacodynamic analysis, the percentage twitch height depression relative to the baseline value obtained just before vecuronium injection was used to calculate the degree of neuromuscular block. Only the first twitch from the train-of-four stimulation was considered to estimate the twitch height depression.

Neuromuscular block was measured continuously and the onset time (time from the end of the vecuronium injection until disappearance of the train of four response), clinical duration (time from the end of injection until 25% recovery of T1), and recovery index (time between 25% and 75% recovery of T1) were determined.

Finally, with the biexponential equation describing vecuronium pharmacokinetics and the observed values of neuromuscular block, the hysteresis curve between predicted plasma concentrations and effect was constructed for each patient. Additional neuromuscular block measurements, other than the percentage block obtained at sampling collection times, were used and associated with predicted plasma concentrations, if needed, to allow a better fitting of the pharmacokinetic-pharmacodynamic model. Care was taken to evenly balance the number of points in the onset and in the recovery phase. The nonparametric link model described by Fuseau and Sheiner [15]was used to minimize the area between the two limbs of the plasma concentration-effect hysteresis and to derive the effect compartment equilibration rate constant for each patient. This parameter was then fitted as such in the sigmoid Emax model described by Holford and Sheiner [16]and used to determine the effect compartment concentration-time profile. The effect concentration and the percentage block observed at each time point was then plotted and a sigmoidal relationship described by the EC⁵⁰(vecuronium effect compartment concentration present at a 50% block), and slope of the sigmoid values was obtained.

The daily carbamazepine dose taken by each patient was noted and total as well as free carbamazepine serum concentrations were measured before surgery using a fluorescence polarization immunoassay. The presence of correlation between the recovery index value and carbamazepine daily dose, carbamazepine serum concentration and free carbamazepine serum concentrations, as well as the presence of correlation between clearance and the three latter variables was verified using simple linear regression.

Demographic data for both groups were compared using Student's t test and chi-square analysis for continuous and categorical data, respectively. To detect a statistical difference in the recovery index between the two groups, a one-sided Mann-Whitney test was used because of a nonnormal distribution of that parameter in the control group. Pharmacokinetic and pharmacodynamic data were compared using a Student's t test for independent samples assuming equal or unequal variance as appropriate (Microsoft Excel 5.0, Redmond, WA). A difference was considered to be significant at a P value of < 0.05. According to previous pharmacodynamic studies [1-3]demonstrating approximately a 50% difference in recovery index, a 50% difference was expected between the carbamazepine and control group EC⁵⁰value. Assuming a standard deviation of 46 ng/ml and an equal sample size per group, the total number of patients needed to detect a clinically significant difference of 50% equivalent to 62 ng/ml in EC sub 50 values between the null and the alternative hypotheses using a two-sample t test with a significance level of 5% and a power level of 80% was equal to 17.26 patients. We assumed a conservative approach and used a sample size of 18 patients.

Demographic data for the two groups as well as the type of surgical procedure undergone are presented in Table 1. The two groups were comparable regarding gender, age, weight, and ASA physical status. One of the patients in the control group was excluded from the overall statistical analysis because he received a 1-g phenytoin infusion for 1 h during the surgical procedure to prevent seizure. Nevertheless, samples from this patient were analyzed to verify if acute administration of phenytoin does effectively potentiate the neuromuscular block induced by vecuronium. [17]Potentiation was observed for that patient.

Concomitant medications received before, during, and after anesthesia until the last blood sample was collected are listed in Table 2. Perioperative medications most often received by patients were 16 mg dexamethasone and 50 g mannitol to control cerebral edema, 1 g oxacillin, and labetolol in doses ranging from 6 to 320 mg.

(Table 3) presents the neuromuscular transmission data for both groups. One patient in each group was excluded from the pharmacodynamic analysis and consequently from the kinetic-dynamic modeling because displacement of the surface electrodes led to unreliable monitoring of the neuromuscular function. Chronic anticonvulsant therapy did not result in prolonged onset times; however, regarding recovery times, twitch height in patients receiving carbamazepine recovered significantly faster than that in patients in the control group. Recovery times were on average shorter by 46% than those for control subjects at each level of recovery (25, 50, 75%, and maximal recovery). The recovery index from T¹25% to T¹75% was significantly decreased in the carbamazepine group as compared to the control group.

(Figure 1) illustrates the kinetic and dynamic raw data obtained in the two groups. Vecuronium observed plasma concentration-time profiles for the control subjects and carbamazepine patients are shown in Figure 1(A and B), respectively. These figures demonstrate the greater population variability in the carbamazepine group.

A two-compartment model could be fitted to the data after the administration of vecuronium. The mean vecuronium plasma concentrations observed at 1 min were 1759 ng/ml and 1392 ng/ml for the control and carbamazepine groups, respectively. In the carbamazepine group, vecuronium mean plasma concentrations decreased more rapidly and were consistently less by at least 43% between 7 and 90 min. At 90 min, mean vecuronium plasma concentrations were of 36 ng/ml and 20 ng/ml for the control and carbamazepine groups, respectively. At 180 min, concentrations could still be quantified in seven of the ten patients in the control group with a mean concentration of 10 ng/ml but only could be measured in one patient in the carbamazepine group. Consequently, the total area under vecuronium plasma concentration curve to time infinity in the carbamazepine group was half the area observed in the control group (11,617+/-1,544 micro gram *symbol* min *symbol* l sup -1 vs. 22,949+/-1,348 micro gram *symbol* min *symbol* l sup -1 presented as mean+/-SEM; P < 0.001). Even though vecuronium metabolites could be quantified by the analytic method for spiked in vitro preparations, they were not detected in the patients' plasma samples.

The pharmacokinetic parameters derived from the two-compartment analysis of vecuronium plasma concentration-time profile for each group appear in Table 4. The clearance was more than doubled in the carbamazepine group, whereas the MRT was halved. No significant changes were observed in the half-lives for both the alpha ([nearly equal] 3 min) and the beta phases ([nearly equal] 28 min) and in the various apparent volumes of distribution (Vss, Vdbeta, and Vc). Only for the B values were differences observed.

Our data also were analyzed using a noncompartmental approach. Within the control and the carbamazepine groups, no statistical difference was observed between clearance and Vdss values as well as the percentage of AUC^t-infinity extrapolated obtained with the two approaches. In the carbamazepine group as compared to the control group, the clearance was statistically significant (10.1+/-1.4 vs. 4.1+/-0.3 ml *symbol* min sup -1 *symbol* kg sup -1 presented as mean+/- SEM; P < 0.003), while the Vdss was not (0.184+/-0.028 vs. 0.136 +/-0.011 L/kg presented as mean+/-SEM; P < 0.148). The total area under the curve in the carbamazepine group was still half the total area observed in the control group (10.601+/-1.512 micro gram *symbol* min *symbol* L sup -1 vs. 21.954+/-1.303 micro gram *symbol* min *symbol* L sup -1 presented as mean+/-SEM; P < 0.001) which was reflected in the percentage of AUC^t-infinity extrapolated in the control versus carbamazepine group (2.4+/- 0.3% versus 4.5+/-0.5% presented as mean+/-SEM; P < 0.002).

Vecuronium observed neuromuscular block-time profiles for the control subjects and carbamazepine patients are shown in Figure 1(C and D), respectively. On average, patients in the carbamazepine block present a more rapid recovery from paralysis induced by vecuronium.

In Figure 2, the sigmoid Emax curve obtained for a typical patient in each group was simulated to appreciate the goodness-of-fit of the pharmacokinetic/pharmacodynamic (PK-PD) modeling. As illustrated by the overlap of the sigmoid Emax curves for patients in the carbamazepine group and in the control group, no statistical difference was observed in the EC⁵⁰(mean EC⁵⁰+/-SEM = 125.0+/-11.2 vs. 154.1+/-19.2 ng/ml in control group). Moreover, no significant changes were observed in any of the pharmacokinetic-pharmacodynamic modeling parameters (Table 5).

In the carbamazepine group, carbamazepine mean daily dose was 1.4 g. Carbamazepine mean therapeutic serum oncentrations measured before surgery was 12.9 micro gram/ml (normal values: 8.0-12.0 micro gram/ml), and 3.1 micro gram/ml for the carbamazepine mean free fraction (normal values: 1.6-2.4 micro gram/ml). No correlation was found between recovery index and carbamazepine daily dose (r²= 0.02), carbamazepine serum concentrations (r²= 0.05), and free carbamazepine serum concentrations (r²= 0.11), as well as between vecuronium clearance and carbamazepine daily dose (r²= 0.07), carbamazepine serum concentrations (r²= 0.10), and free carbamazepine serum concentrations (r²= 0.02).

This study describes the mechanisms underlying the resistance to vecuronium-induced neuromuscular blockade in epileptic patients taking carbamazepine on a chronic basis. Our major finding was a twofold increase in clearance in the carbamazepine-treated group, which provides evidence of a pharmacokinetic origin to the shorter recovery observed in theses patients. Clinically, in patients in the carbamazepine group twitch height recovered twice as fast when compared with control subjects. Onset time was not changed. No difference in the EC⁵⁰values was found between the two groups.

The clearance in the control group was consistent with published values of 3.0-5.6 obtained when vecuronium is given without prior administration of succinylcholine. [18-20]However, in our control group, Vdss and MRT values were consistently smaller than those obtained by our laboratory in a previous study in anesthetized patients undergoing elective surgeries. [18]These differences can be explained in part by different patient population. In the current study, patients underwent a neurosurgical procedure that resulted in four of ten patients receiving intravenous perioperative diuretics such as furosemide and infusions of mannitol. Administration of these diuretic agents may have resulted in an increased diuresis and consequently a smaller extracellular volume or Vdss. To a lesser extent, diuresis may have also increased, but not significantly, the clearance in the control group (3.8+/-0.3 vs. 3.5+/-0.2 ml *symbol* min sup -1 *symbol* kg sup -118) and, consequently, a shorter MRT was observed. The presence of mannitol in the plasma samples was often associated with a loss in reactivity of the electrochemical cell and hence, a decreased sensitivity of the analytical method ensued. Nevertheless, since only 5% of the total AUC was extrapolated, this possible loss of sensitivity would not have significantly altered the MRT value.

In patients receiving carbamazepine on a long-term basis, vecuronium clearance was increased twofold. Several approaches to the determination of vecuronium kinetic parameters could have been selected. A non-parametric model often is ideal because no prior assumptions need to be made about the pharmacokinetic model or the plasma concentration effect function. However, frequent sampling must be done between 0 and 2 min for the onset phase to be described appropriately. Failure to do so will result in an underestimation of the AUC covering the first 2 min and an over estimation of the kinetic parameters derived. Ducharme et al. [18]demonstrated that similar vecuronium kinetic and dynamic parameters could be obtained when using a compartmental analysis associated with limited sampling or a noncompartmental analysis associated with continuous sampling over the first 2 min after injection. Consequently, the compartmental approach was chosen because it required less frequent sampling and was therefore more easily applicable in a clinical setting. Nevertheless, our results were verified using both a compartmental and a noncompartmental approach. No bias was introduced by the pharmacokinetic model chosen because similar conclusions were drawn with both methods. This provides important confirmatory evidence of our fundamental findings, which clearly suggest a pharmacokinetic basis to the resistance.

Increased hepatobiliary excretion that may be associated with increased metabolism due to enzymatic induction is a plausible underlying mechanism for the increased clearance in the patients receiving carbamazepine. In humans, hepatic uptake and biliary excretion account for 70% of the elimination of a vecuronium dose, the kidney only playing a minor role. [21,22]Bencini et al. calculated that, in humans, only half an hour after injection, as much as 80% of a vecuronium dose can be found in the liver with 40% excreted mainly unchanged in the bile. [21,23]This suggests that the liver may act as a reservoir for vecuronium. [21,24,25]Vecuronium's uptake by the liver proceeds at a faster rate than its excretion into the bile; which explains the large amounts of vecuronium found in the liver as well as the rapid decline in plasma vecuronium concentrations.

In patients receiving enzyme-inducing antiepileptic medication for at least 1 yr, Pirttiaho and Sotaniemi observed a liver weight (kg/body weight) 30% greater than that of normal patients and accompanied by a proportional increase in liver blood flow. [26,27]This hepatomegaly would explain the increased clearance observed with vecuronium. Furthermore, the decreased plasma concentrations as well as the twofold decrease in B value in the carbamazepine group could be attributed to a larger liver reservoir capacity. Finally, not only would hepatomegaly explain the increased clearance, but it could also explain why resistance to the effect of nondepolarizing neuromuscular relaxants is observed in patients receiving either carbamazepine or phenytoin. The two drugs are structurally unrelated antiepiletics, yet both are strong hepatic enzyme inducers.

Ornstein et al., in evaluating the resistance to metocurine in patients who had been receiving phenytoin preoperatively for at least 2 weeks, compared the pharmacokinetics and pharmacodynamics of metocurine in the phenytoin group with those in the control group. They did not observe differences in metocurine's pharmacokinetics. [9]Unlike vecuronium, metocurine's elimination relies chiefly on renal elimination. Consequently, the presence of hepatomegaly would not alter metocurine's pharmacokinetics. [28].

In patients receiving carbamazepine on a chronic basis, a shift to the right of the ED⁵⁰was reported in a study by Whalley et al. [3]It can be hypothesized that when small doses of vecuronium such as 0.01 mg/kg are administered, plasma vecuronium concentrations decrease to less than pharmacologically active concentrations before the late distribution phase is reached and the neuromuscular effect is hence governed by distribution rather than elimination. However, after a larger dose of 0.1 mg/kg, paralysis probably occurs during the elimination phase and clinical duration and recovery index will be the parameters affected. [29]In patients receiving carbamazepine, the recovery index from T¹25% to T¹75% was decreased (7.6+/-1.2 min. vs. 21.9 +/-6.5 min. in the control group). Interestingly, similar shorter recovery profiles have been reported for patients receiving carbamazepine and other muscle relaxants. [1,2].

The presence of enzymatic induction on vecuronium metabolism by the antiepileptic drugs could have been strongly suggested by an increase in the hydroxyl metabolites of vecuronium. However, quantifiable concentrations of these metabolites were not observed in our plasma samples. Consequently, the twofold increase in clearance could not be accounted for by an increased hepatic metabolism due to enzymatic induction. In fact, there is general agreement [30,31]that after clinical doses of vecuronium, the 3-OH-derivative is either undetectable or represents only 5% to 10% of the parent compound concentration, while circulating concentrations of the 17-OH or the 3-OH, 17-OH derivatives are never observed in patients. It follows that vecuronium's metabolism is either minimal or that rapid hepatic uptake of these metabolites occurs and obscures the presence of enzyme induction. If the presence of interaction relies in part on enzyme induction and subsequent hepatic protein synthesis, resistance to neuromuscular blockers should not be observed in patients receiving an anticonvulsant with minor enzyme-inhibiting properties such as valproic acid. [32]However, previously documented clinical evidence remains unavailable.

Increased binding of the neuromuscular relaxants to alpha-1-acid glycoproteins (AAG) also has been suggested as a contributing factor to the increased dose requirement for muscle relaxants or accelerated recovery in patients receiving enzyme-inducing anticonvulsants. [4,8]However, in a study conducted in rats, Kim et al. documented a tenfold increase in AAG proteins after 14 days of phenytoin therapy, but noted only a 30% increase in metocurine protein binding and concluded that altered binding alone did not explain the interaction. [8]It should be kept in mind that, even though an increase in AAG proteins may be seen in humans after long-term anticonvulsant therapy, in the case of neurosurgical procedures, patients receive many concomitant medications such as dexamethasone or oxacillin, which are highly bound to plasma proteins and may counteract effects caused by increased AAG binding. [8].

Using this pharmacokinetic-pharmacodynamic modeling approach, the obtained EC⁵⁰values obtained for both groups were consistent with previously published data [18]and no statistically significant difference in the EC⁵⁰could be found between the two groups. Furthermore, we can expect no changes in EC⁹⁰values because no differences between the slope of the sigmoid values were seen. Ornstein et al. had suggested that the interaction observed with metocurine be of a pharmacodynamic nature. In light of the current results, the possibility of a concurrent pharmacodynamic alteration still cannot be discounted. In the presence of an increase in AAG, [8]for a similar EC⁵⁰, vecuronium's clinical effect would be decreased in the carbamazepine treated group. Greater knowledge of the importance of protein drug binding should be acquired to give to the EC⁵⁰value a meaningful interpretation.

In conclusion, this study represents the first attempt to use compartmental pharmacokinetic-pharmacodynamic modeling to investigate the interaction between carbamazepine and vecuronium. The current data suggest that for muscle relaxants subject to hepatobiliary elimination, the importance of pharmacokinetic mechanisms may have been underestimated until now; nevertheless, pharmacodynamic mechanisms may also be involved in the observed resistance and these mechanisms should not be discounted until greater knowledge of the importance of drug binding is gained.