Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

The azole antifungal agent ketoconazole impairs clearance and enhances the central nervous system effects of several benzodiazepines metabolized by CYP3A isoforms, including triazolam, midazolam, and alprazolam (von Moltke et al., 1996a; Greenblatt et al., 1998b; Schmider et al., 1999; Tsunoda et al., 1999; Yuan et al., 1999; Venkatakrishnan et al., 2000, 2001). For orally administered benzodiazepine derivatives, such as triazolam and midazolam, that ordinarily undergo extensive hepatic and gastrointestinal presystemic extraction, coadministration of usual therapeutic doses of ketoconazole may increase area under the plasma concentration curve by 10-fold or more (von Moltke et al., 1996b; Greenblatt et al., 1998b; Tsunoda et al., 1999; Yuan et al., 1999; Venkatakrishnan et al., 2000). For low extraction benzodiazepines such as alprazolam, ketoconazole also reduces apparent oral clearance, although the magnitude of the interaction is less dramatic (Greenblatt et al., 1998b; Schmider et al., 1999). The capacity of ketoconazole to impair biotransformation of these and other CYP3A substrates in vitro also is well established (Thummel and Wilkinson, 1998; Venkatakrishnan et al., 2000, 2001; von Moltke et al., 1994, 1996a,b). Ketoconazole is a highly potent CYP3A inhibitor, with inhibition constant (K_i) values generally falling in the nanomolar range.

A clinical drug interaction study involving triazolam and ketoconazole indicated that the enhanced benzodiazepine agonist effects of triazolam due to coadministration of ketoconazole were less than would be anticipated based on the elevation of plasma triazolam levels (von Moltke et al., 1996a). Experimental observations indicate that ketoconazole binds to the benzodiazepine receptor and may act as a functional antagonist (Fahey et al., 1998). This unique property of ketoconazole indicates that ketoconazole influences not only pharmacokinetics but also pharmacodynamics of benzodiazepines.

The present study evaluated an experimental paradigm to assess the potential applicability of an experimental pharmacokinetic-pharmacodynamic model for study of complex drug interactions of this type. We also characterized the P450 isoforms contributing to midazolam biotransformation in the rat, and evaluated the extent to which in vitro data on ketoconazole inhibition of midazolam biotransformation in vitro is predictive of actual pharmacokinetic interactions observed in vivo.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Animals. Male Sprague-Dawley rats (n = 4) weighing 300 to 350 g were housed individually under a 12-h light/dark cycle (lights on at 6:00 AM). Food and water were provided ad libitum, except on the trial days described below. Details of animal housing and maintenance, surgical procedures, and general study methodology have been described previously (Kotegawa et al., 1998, 1999; Laurijssens and Greenblatt, 2002).

Experimental Design. A four-way crossover design was used. The four treatment conditions were as follows: 1) 5 mg/kg midazolam intravenously, with intraperitoneal ketoconazole vehicle administered as described below; 2) 5 mg/kg midazolam intravenously, with intraperitoneal ketoconazole; 3) 15 mg/kg midazolam intragastrically, with intraperitoneal ketoconazole vehicle; and 4) 15 mg/kg midazolam intragastrically, with intraperitoneal ketoconazole. The sequence of these four trials was randomized. The washout period was 8 days if ketoconazole was administered in the previous trial, and 4 days if vehicle was administered in the previous trial. Seven days after completion of the four trials, rats were sacrificed and the liver and intestine were removed.

Surgery. Five days before the first trial, the surgery was performed under ketamine/xylazine anesthesia (Kotegawa et al., 1998, 1999; Laurijssens and Greenblatt, 2002). Two silicon rubber cannulas (Silastic; Dow Corning Co., Lansing, MI) were implanted. One cannula (length, 20 cm; internal diameter, 0.5 mm; and external diameter, 0.94 mm) was implanted in the right jugular vein, and the other, which was made by combining two different sizes of silicon tube (length, 15 and 5 cm; internal diameter, 0.5 and 0.3 mm; and external diameter, 0.94 and 0.63 mm), was implanted in the left carotid artery. The distal ends of both cannulas were tunneled subcutaneously, and exited between the ears. The venous cannula, used for drug administration, was positioned in the vena cava, and the arterial cannula, used for blood sampling, was positioned in the descending aorta. Part of a plastic tube with a cap was fixed on the skull using cranioplastic cement and the tunneled cannulas were fixed in the tube. Five EEG electrodes were also implanted onto the skull as follows: 11 mm anterior and ± 2.5 mm lateral to  (Fl and Fr), 3 mm anterior and ± 3.5 mm lateral to  (Cl and Cr), and  (reference). The electrode consisted of a stainless steel screw connected to a male amphenol pin with about 1 cm of insulted wire. The five electrodes placed to the skull were fixed using cranioplastic cement. The ends of the electrodes were connected to a miniature connector with EEG cables on the trial day. The electrodes were protected using an identical miniature connector without EEG cables during washout periods.

Pharmacokinetic and Pharmacodynamic Study in Vivo. Midazolam solution was prepared as 5 mg/ml midazolam base for intravenous administration, or 25 mg/ml midazolam base for intragastric administration, in 0.9% saline, pH 3.5. The ketoconazole solution was 10 mg/ml dissolved in polyethylene glycol 400. On the trial day, rats received an intravenous bolus dose of 5 mg/kg midazolam and an intragastric vehicle, or an intragastric dose of 15 mg/kg midazolam and an intravenous vehicle according to the predetermined schedule. Ketoconazole dosage was an intraperitoneal dose of 10 mg/kg 30 min before midazolam dosage, which was followed by 5 mg/kg given 180 min after the initial dose. This schedule maintains serum ketoconazole concentrations at 2 µg/ml or higher until at least 500 min after administration (Kotegawa et al., 1999). Blood samples for midazolam quantitation were collected from the arterial cannula up to 480 min after midazolam administration. Blood samples were centrifuged and the separated serum was stored at 20°C until the time of assay. Serum midazolam concentrations were determined by gas chromatography with electron capture detection (Arendt et al., 1984).

Liver and Intestinal Microsome Preparation. Rat liver microsomes were prepared as described previously (von Moltke et al., 1994, 1996a,b; Cotreau et al., 2000; Perloff et al., 2000). Small intestinal epithelial cell isolation and subsequent microsomal preparation were described in detail previously (Cotreau et al., 2000). Prepared liver and intestinal microsomes were stored at 80°C until use.

Incubation Procedure and Midazolam Metabolite Assay. The microsomal incubation procedure was similar to that described previously (von Moltke et al., 1996b; Cotreau et al., 2000; Perloff et al., 2000). The protein content and reaction time were predetermined based on linearity between microsomal protein concentration (up to 0.5 mg/ml) and the reaction time (up to 13 min) versus metabolite formation rate. Incubation mixtures contained 50 mM phosphate buffer, 5 mM Mg²⁺, 0.5 mM NADP⁺, and an isocitrate/isocitric dehydrogenase-regenerating system. Incubations were performed at 37°C with 0 to 300 µM midazolam and 0, 1, or 2 µM ketoconazole. Reactions were initiated by addition of microsomal protein (approximately 0.1 mg/tube; final volume, 250 µl). The incubation time was 7 min for liver microsomes or 10 min for intestinal microsomes. The reactions were stopped by cooling on ice and adding 100 µl of acetonitrile. Phenacetin was added as the internal standard. The incubation mixture was centrifuged, and the supernatant was subjected to HPLC analysis. The HPLC assay for -hydroxy-midazolam (-OH-midazolam) and 4-hydroxy-midazolam (4-OH-midazolam) was performed as described previously (von Moltke et al., 1996b; Cotreau et al., 2000; Perloff et al., 2000; Warrington et al., 2000).

Protein Binding. An equilibrium dialysis technique was used to examine the effect of ketoconazole on midazolam binding to serum proteins (Moschitto and Greenblatt, 1983; Laurijssens and Greenblatt, 2002). A 0.45-ml aliquot of rat serum containing 1 µg/ml midazolam and 0 to 10 µg/ml ketoconazole was placed in cellulose dialysis tubing that was sealed and submerged in 1.2 ml of phosphate buffer, pH 7.4. Samples were incubated at 37°C for 20 h. Midazolam concentrations in the dialyzed serum and dialysate were measured by gas chromatography (Arendt et al., 1984). Free fraction was calculated as the ratio of these concentrations.

Data Analysis. After intravenous midazolam, data points were fitted to a linear sum of two exponential terms. The fitted function was used to calculate elimination half-life, volume of distribution using the area method, total area under the serum concentration curve (AUC), and total clearance. After oral midazolam, the slope of the terminal elimination phase was used to calculate the elimination half-life. Area under the curve up to the final detectable concentration was measured using the linear trapezoidal method and extrapolated to infinity, yielding total AUC. Apparent clearance after intragastric administration was calculated as dose divided by total AUC. Systemic bioavailability of intragastric midazolam (F) was calculated as total AUC after intragastric administration (with or without ketoconazole) divided by total AUC for the corresponding intravenous trial, after normalization for differences in dosage (Tsunoda et al., 1999).

(1)

(2)

(3)

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

Body Weight Change. Body weight (a general index of the animals' health) increased throughout the study, except for the day of surgery or experimental trials (Fig. 1). One rat showed a weight reduction 2 days after the last trial. Pharmacokinetic and pharmacodynamic data obtained from this animal were used in the analysis, but liver and intestinal microsomes were not prepared from this animal. Additional microsomes were obtained from another rat.

View larger version (17K): [in this window] [in a new window]   Fig. 1.   Change in body weight of rats (mean ± S.D.; n = 4). One rat was eliminated 2 days after trial 4 because of poor health. Closed circles indicate the mean of three rats.

Pharmacokinetics of Midazolam in Vivo. Coadministration of ketoconazole caused a significant increase in AUC and decrease in clearance of intravenous midazolam (Table 1; Fig. 2). The mean (±S.E.) ratio (value with ketoconazole divided by control value) was 1.45 (± 0.56) for AUC and 0.69 (± 0.26) for clearance. The effect of ketoconazole was greater when midazolam was administered intragastrically. Ketoconazole caused a significant increase in C_max and total AUC and a significant decrease in apparent oral clearance (Table 1; Fig. 2). Mean ratios were 6.5 (± 2.2) for AUC and 0.15 (± 0.05) for oral clearance. Midazolam oral bioavailability (F) increased from 0.11 in the control condition to 0.36 with ketoconazole (Table 1). Ketoconazole produced no significant change in midazolam elimination half-life either after intragastric or intravenous midazolam administration (Table 1), although half-life tended to be longer after intragastric as opposed to intravenous midazolam dosage.

View larger version (13K): [in this window] [in a new window]   Fig. 2.   Mean (± S.E.) serum midazolam concentrations after an intravenous bolus dose of 5 mg/kg midazolam (left) or an intragastric dose of 15 mg/kg (right), with and without coadministration of ketoconazole.

View larger version (10K): [in this window] [in a new window]   Fig. 3.   Effect of ketoconazole on midazolam binding to rat serum protein (mean ± S.E.; n = 4).

View larger version (11K): [in this window] [in a new window]   Fig. 4.   Relation of hepatic blood flow to predicted FH and FG after intragastric administration of midazolam in rats. Left, control condition, without ketoconazole. Assumptions are: net bioavailability (F) = 0.11 and intravenous clearance = 79 ml/min/kg. Right, with coadministration of ketoconazole, F = 0.36, and intravenous clearance = 55 ml/min/kg.

Pharmacodynamics of Midazolam in Vivo. Administration of midazolam produced characteristic increases in -amplitude on the EEG (Fig. 5). After i.v. dosage, ketoconazole had no detectable influence on -amplitude; mean area under the EEG effect versus time curve was essentially identical in the control condition and with ketoconazole. After intragastric midazolam, EEG -amplitude was enhanced by an average of 11% (Fig. 5), but the difference did not reach significance.

View larger version (14K): [in this window] [in a new window]   Fig. 5.   The percent change over baseline in EEG amplitude falling in the -frequency range after an intravenous bolus dose of 5 mg/kg midazolam (left) or an intragastric dose of 15 mg/kg midazolam (right), with and without coadministration of ketoconazole. Each point is the mean value for the four animals at the corresponding time.

View larger version (17K): [in this window] [in a new window]   Fig. 6.   Relationship between midazolam concentrations at the hypothetical effect site and EEG effect in a representative animal after an intravenous bolus dose of 5 mg/kg midazolam (left) or an intragastric dose of 15 mg/kg midazolam (right), with and without coadministration of ketoconazole. The lines represent fitted functions consistent with a sigmoid Emax model (eq. 3).

Midazolam Metabolism in Vitro. Midazolam was biotransformed to -OH-midazolam and 4-OH-midazolam by liver microsomes (Table 3; Fig. 7) and intestinal microsomes (Table 3; Fig. 8). In liver microsomes, K_m values for both pathways were similar, and the 4-OH pathway accounted for the majority of intrinsic clearance. Ketoconazole inhibited midazolam biotransformation via both pathways by an apparently competitive mechanism, with mean K_i values of 0.65 µM for the -OH pathway and 0.14 for the 4-OH pathway. In intestinal microsomes, K_m values differed from those in liver, with the -OH pathway having a lower K_m value (mean 6 µM) than that for the 4-OH pathway (55 µM). The -OH pathway accounted for the majority of intrinsic clearance in intestinal microsomes. K_i values for ketoconazole inhibition were in the nanomolar range.

View larger version (16K): [in this window] [in a new window]   Fig. 7.   In vitro formation of midazolam metabolites from midazolam by a representative preparation of rat liver microsomes. Solid lines (control condition, without ketoconazole) are functions consistent with Michaelis-Menten kinetics and uncompetitive substrate inhibition. Dashed lines (with coaddition of ketoconazole) are consistent with competitive inhibition.

View larger version (13K): [in this window] [in a new window]   Fig. 8.   In vitro formation of midazolam metabolites from midazolam by a representative preparation of rat intestinal microsomes. Solid lines (control) are functions consistent with Michaelis-Menten kinetics. Dashed lines (with ketoconazole) are consistent with competitive inhibition.

View larger version (23K): [in this window] [in a new window]   Fig. 9.   Formation of 4-OH-midazolam (above) and -OH-midazolam (below) from 50 µM midazolam by heterologously expressed rat P450s, by the human lymphoblastoid cell and insect cell expression vectors (L. vector and S. vector, respectively), by human and rat liver microsomes (HLMs and RLMs), respectively, and by negative control incubations from which cofactors, microsomal protein, or substrate (midazolam) was successively omitted. For all cDNA-expressed enzymes, except CYP2E1, 50 pmol of cytochrome P450/ml was used. For CYP2E1 and vector controls, 1 mg/ml of protein was used. The x-axis indicates the quantity of metabolite formed during a 7-min incubation.

View larger version (14K): [in this window] [in a new window]   Fig. 10.   Relation of ketoconazole concentration to net velocity of midazolam hydroxylation in vitro, expressed as a fraction of the control velocity with no ketoconazole present. The closed circle represents the actual in vivo decrement in midazolam clearance (clearance with ketoconazole divided by clearance in the control condition; Table 1) after intragastric administration; this observed clearance ratio relative to control (0.15) corresponds to a hypothetical in vitro enzyme-available concentration of 1.13 µM (0.6 µg/ml). The open circle indicates the predicted clearance ratio (0.06) corresponding to a total (free plus bound) serum ketoconazole concentration in vivo of 2 µg/ml (3.76 µM). The open squares indicate the predicted clearance ratios (0.56 and 0.83) corresponding to anticipated unbound serum ketoconazole concentrations in vivo of 0.14 or 0.04 µM, respectively.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Consistent with previous studies (Mandema and Danhof, 1992; Lau et al., 1996; Higashikawa et al., 1999a,b,c; Laurijssens and Greenblatt, 2002), clearance of intravenously administered midazolam in rats was high, with a mean control clearance value (without inhibitor) of 79 ml/min/kg. Intragastric midazolam had a mean net bioavailability of only 11%. Partitioning the net F value into hepatic and enteric components requires an assumed value of hepatic flow, and the specific estimates of F_H and F_G are highly dependent on the Q_H value entered into the model. Small changes in Q_H within the range of 90 to 110 ml/min/kg yield large changes in F_H and F_G (Fig. 4). Nonetheless, the general conclusions are consistent with previous studies in the rat in that clearance of intravenous midazolam is largely flow-dependent, and incomplete bioavailability of oral midazolam is principally determined by hepatic as opposed to enteric presystemic extraction. These relationships are only partly consistent with the pattern of midazolam kinetics in humans, in whom typical values of intravenous clearance (5-12 ml/min/kg) are less than 50% of hepatic blood flow, oral bioavailability averages 25 to 40%, and intestinal metabolism contributes in a major way to incomplete oral bioavailability (Greenblatt et al., 1984; Gorski et al., 1998; Thummel and Wilkinson, 1998; Tsunoda et al., 1999).

The dosage regimen of ketoconazole used in the present study has been verified to produce serum ketoconazole concentrations consistently in excess of 2 µg/ml (Kotegawa et al., 1999), which is similar to the range encountered with usual therapeutic doses of ketoconazole in humans (Venkatakrishnan et al., 2000). Clearance of intravenous midazolam was reduced by ketoconazole to 69% of control values, whereas apparent clearance of intragastric midazolam was reduced to 15% of control. Again, this is consistent with the high clearance of midazolam in the rat, such that ketoconazole would inhibit the component of intravenous clearance that is not flow-dependent, whereas ketoconazole would have a much greater effect on oral clearance. In humans, clearance of intravenous midazolam is 35 to 40% of hepatic blood flow, and ketoconazole reduced intravenous clearance to 20% of control values (Tsunoda et al., 1999). Clearance of oral midazolam in humans was reduced to about 6% of control values. Thus, species differences in the pharmacokinetics of midazolam, the relation of clearance to hepatic blood flow, and the relative contributions of hepatic and enteric sites to net presystemic extraction, are evident as predictable difference in the response to chemical inhibition by ketoconazole.

In agreement with previous studies from our laboratory and elsewhere (Mandema and Danhof, 1992; Laurijssens and Greenblatt 1996, 2002), administration of midazolam to the rat produced increases in EEG -amplitude consistent with its benzodiazepine agonist properties. Kinetic-dynamic modeling of the data required inclusion of a hypothetical effect site, with apparent equilibration half-life values estimated in the range of 5 to 15 min, as described with intravenous midazolam administration in humans (Mandema and Danhof, 1992; Laurijssens and Greenblatt, 1996). We observed that coadministration of ketoconazole had no significant effect on the estimated E_max for the EEG effect parameter, but caused a significant rightward shift in the relationship between midazolam hypothetical effect site concentration and EEG effect. This was evident as an increase in the estimated EC₅₀ values. A similar phenomenon was reported in a clinical study of triazolam and ketoconazole in humans (von Moltke et al., 1996a) and in an experimental model (Fahey et al., 1998). This may be explained by the property of ketoconazole as a neutral ligand at the benzodiazepine receptor, acting functionally as an antagonist and shifting the concentration-response relationship (Fahey et al., 1998). It should be noted that ketoconazole alone has no detectable effect on the EEG, either in this experimental model (Kotegawa et al., 1999) or in humans (von Moltke et al., 1996a; Greenblatt et al., 1998b).

Midazolam was biotransformed to two principal metabolites, -OH-midazolam and 4-OH-midazolam, by rat liver and intestinal microsomes. In liver microsomes, K_m values for both pathways were in the range of 20 to 25 µM, and the 4-OH-midazolam dominated in terms of contribution to net intrinsic clearance. These results are similar to previous studies of rat liver microsomes (Ghosal et al., 1996), but differ from kinetic parameters in human liver microsomes (Kronbach et al., 1989; Gorski et al., 1994; Ghosal et al., 1996; von Moltke et al., 1996b; Perloff et al., 2000). Likewise, rat liver was less susceptible to inhibition of midazolam biotransformation by ketoconazole compared with human liver microsomes (Ghosal et al., 1996). Testosterone hydroxylation also was less susceptible to ketoconazole inhibition in rat as opposed to human liver microsomes (Eagling et al., 1998). Midazolam is presumably metabolized almost exclusively by CYP3A4 and 3A5 in humans. In rats, however, 4-OH-midazolam formation is largely mediated by CYP3A1 and 3A2, but other P450 isoforms of the CYP2C subfamily contribute substantially to formation of -OH-midazolam (Fig. 9). This is consistent with the findings of Kobayashi et al. (2002). As in many previous studies (von Moltke et al., 1994, 1996a,b; Yamano et al., 1999a; Perloff et al., 2000; Venkatakrishnan et al., 2000), ketoconazole inhibition was consistent with a predominantly competitive mechanism, although noncompetitive inhibition has also been described in the literature.

Kinetic parameters for midazolam hydroxylation by rat intestinal microsomes differed from the profile in liver microsomes in terms of K_m values, the relative contributions of the two pathways to intrinsic clearance, and susceptibility to inhibition by ketoconazole. It is possible that rat intestine may express different CYP3A isoforms compared with liver (Gushchin et al., 1999), thereby leading to kinetic differences. Expression of CYP2C isoforms that contribute importantly to midazolam -hydroxylation also may differ between hepatic and enteric sites.

Using conventional in vitro-in vivo scaling techniques (von Moltke et al., 1995, 1998; Venkatakrishnan et al., 2000, 2001), we combined in vitro ketoconazole K_i values for rat liver microsomes together with anticipated in vivo total serum ketoconazole concentrations (2 µg/ml) based on a previous study of similar design (Kotegawa et al., 1999) to evaluate the actual reduction in clearance of intragastric midazolam caused by ketoconazole in relation to what would be predicted from the in vitro data. Assuming an in vitro inhibitor concentration equal to the in vivo total serum concentration (2 µg/ml; 3.76 µM), the in vitro reaction velocity would be predicted to be reduced to 6% of control. In the actual in vivo study, clearance of intragastric midazolam was reduced to 15% of control. In contrast, anticipated in vivo unbound serum midazolam concentrations (based on two literature estimates of ketoconazole free fraction in the rat) yielded predicted in vivo reaction velocity ratios of 56 and 83% of control, respectively. Therefore, use of total serum ketoconazole concentrations for purposes of scaling yields predictions that slightly overestimate the observed extent of interaction. However, use of unbound serum ketoconazole levels grossly underestimates the actual interaction, apparently attributable to underestimation of the in vivo enzyme-available ketoconazole concentration by a factor falling between 8- and 32-fold. In clinical and in vitro studies of the interaction of zolpidem and ketoconazole in humans, unbound plasma ketoconazole concentrations likewise greatly underestimated the observed clinical interaction, attributable to a 23-fold underestimation of the enzyme-available ketoconazole concentration in vivo (Greenblatt et al., 1998a; von Moltke et al., 1999). These findings are consistent with many previous studies indicating that it is not generally valid to equate unbound serum concentrations of inhibitor to the concentrations available at the metabolic enzyme site (von Moltke et al., 1998; Yamano et al., 1999a,b; Venkatakrishnan et al., 2000, 2001).

The model described herein may be of value as an experimental paradigm to screen for pharmacokinetic drug interactions involving benzodiazepine receptor agonists and antagonists. The clinical applicability of the model is limited by species differences in patterns of metabolic biotransformation and systemic pharmacokinetics, as well as susceptibility to inhibition by chemical inhibitors.