PREDICTION OF THE IN VIVO INTERACTION BETWEEN MIDAZOLAM AND MACROLIDES BASED ON IN VITRO STUDIES USING HUMAN LIVER MICROSOMES

Transcription

1 /03/ $7.00 DRUG METABOLISM AND DISPOSITION Vol. 31, No. 7 Copyright 2003 by The American Society for Pharmacology and Experimental Therapeutics 924/ DMD 31: , 2003 Printed in U.S.A. PREDICTION OF THE IN VIVO INTERACTION BETWEEN MIDAZOLAM AND MACROLIDES BASED ON IN VITRO STUDIES USING HUMAN LIVER MICROSOMES KIYOMI ITO, KANAKO OGIHARA, SHIN-ICHI KANAMITSU, AND TOMOO ITOH School of Pharmaceutical Sciences, Kitasato University, Shirokane, Minato-ku, Tokyo, Japan (Received September 3, 2002; accepted March 28, 2003) This article is available online at ABSTRACT: Clinical studies have revealed that plasma concentrations of midazolam after oral administration are greatly increased by coadministration of erythromycin and clarithromycin, whereas azithromycin has little effect on midazolam concentrations. Several macrolide antibiotics are known to be mechanism-based inhibitors of CYP3A, a cytochrome P450 isoform responsible for midazolam hydroxylation. The aim of the present study was to quantitatively predict in vivo drug interactions in humans involving macrolide antibiotics with different inhibitory potencies based on in vitro studies. - and 4-Hydroxylation of midazolam by human liver microsomes were evaluated as CYP3A-mediated metabolic reactions, and the effect of preincubation with macrolides was examined. The hydroxylation of midazolam was inhibited in a time- and concentration-dependent manner following preincubation with macrolides in the presence of NADPH, whereas almost no inhibition was observed without preincubation. The kinetic parameters Macrolide antibiotics have now been used to treat a variety of infectious diseases for several decades and are often coadministered with other drugs. Like ketoconazole, erythromycin has been reported to cause severe cardiac toxicity with terfenadine when coadministered in humans (Food and Drug Administration, 1990). Honig et al. (1994) have reported that the plasma concentrations of terfenadine, which is undetectable when given alone (60 mg b.i.d. for 7 days), increased to detectable levels following coadministration of erythromycin (500 mg t.i.d. for 7 days). This interaction has been shown to be caused by the inhibitory effect of erythromycin on CYP3A, a cytochrome P450 isoform responsible for terfenadine metabolism (Jurima-Romet et al., 1994). The degree of inhibition of CYP3A varies among the macrolides: the area under the concentration-time curve (AUC) of orally administered midazolam, a substrate of CYP3A, was reported to increase by a factor of 3.8 or 3.6 after pretreatment with erythromycin (500 mg t.i.d. for 5 days) (Zimmermann et al., 1996) or clarithromycin (250 mg This work was supported in part by a Kitasato University Research Grant for Young Researchers. 1 Abbreviations used are: AUC, area under the concentration-time curve; o.d., once daily; P450, cytochrome P450; HPLC, high-performance liquid chromatography. Address correspondence to: Dr. Tomoo Itoh, School of Pharmaceutical Sciences, Kitasato University, Shirokane, Minato-ku, Tokyo , Japan. itoht@pharm.kitasato-u.ac.jp for enzyme inactivation (K app and k inact ) involved in midazolam -hydroxylation were 12.6 M and min 1, respectively, for erythromycin, 41.4 M and min 1, respectively, for clarithromycin, and 623 M and min 1, respectively, for azithromycin. Similar results were obtained for the 4-hydroxylation pathway. These parameters and the reported pharmacokinetic parameters of midazolam and macrolides were then used to simulate in vivo interactions based on a physiological flow model. The area under the concentration-time curve (AUC) of midazolam after oral administration was predicted to increase 2.9- or 3.0-fold following pretreatment with erythromycin (500 mg t.i.d. for 5 or 6 days, respectively) and 2.1- or 2.5-fold by clarithromycin (250 mg b.i.d. for 5 days or 500 mg b.i.d. for 7 days, respectively), whereas azithromycin (500 mg o.d. for 3 days) was predicted to have little effect on midazolam AUC. These results agreed well with the reported in vivo observations. b.i.d. for 5 days) (Yeates et al., 1996), respectively. On the other hand, a relatively small increase, by a factor of 1.5 and 1.2, has been reported in the case of pretreatment with roxithromycin (300 mg o.d. for 6 days) (Backman et al., 1994) and azithromycin (500 mg o.d. for 3 days) (Yeates et al., 1996), respectively. These interactions are based on a mechanism-based inhibition (Silverman, 1988), which differs from competitive or noncompetitive inhibition. CYP3A demethylates the macrolide to a nitrosoalkane which then forms a stable, inactive complex with P450 (Periti et al., 1992). In such a case, the inhibitory effect remains after elimination of the inhibitor from plasma or tissues, which may lead to more serious toxicity compared with the case of reversible inhibition. We have already succeeded in making quantitative predictions of in vivo 5-fluorouracil/sorivudine and triazolam/erythromycin interactions, both involving mechanism-based inhibition of metabolic enzymes, based on a physiologically based pharmacokinetic model and in vitro data (Kanamitsu et al., 2000a,b; Ito et al., 1998). In the present study, using a similar methodology, an attempt was made to predict the degree of in vivo drug interactions in humans involving macrolides with different inhibitory potencies based on in vitro metabolic inhibition studies. Materials and Methods Chemicals and Reagents. Midazolam, -hydroxy (OH) midazolam, and 4-OH midazolam were generously donated by Nippon Roche K.K. (Tokyo, Japan), etizolam by Welfide Corp. (Osaka, Japan), clarithromycin by Taisho Pharmaceutical Co., Ltd. (Tokyo, Japan), and azithromycin by Pfizer, Inc. 945

2 946 ITO ET AL. (Groton, CT). Erythromycin and EDTA-2Na were purchased from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). NADP, glucose 6-phosphate and glucose-6-phosphate dehydrogenase were obtained from Roche Diagnostics (Mannheim, Germany). Acetonitrile, methanol, and other reagents of analytical grade were purchased from Kanto Chemical Co. (Tokyo, Japan). Pooled human liver microsomes (H161) were a gift from BD Gentest (Woburn, MA). Midazolam Metabolism by Pooled Human Liver Microsomes. After 5 min of preincubation, at 37 C, of a reaction mixture (0.72 ml) consisting of 0.1 mg/ml human liver microsomes and an NADPH-generating system (0.33 mm NADP, 8 mm glucose 6-phosphate, 0.1 U/ml glucose-6-phosphate dehydrogenase, 6 mm MgCl 2 ) in 100 mm potassium phosphate buffer (ph 7.4) containing 0.1 mm EDTA, enzyme reactions were initiated by adding 80 l of midazolam in 20% acetone solution. After incubation at 37 C in a shaking water bath for 3 min, the reaction was terminated by transferring the 600- l aliquot to another tube containing 800 l of ice-cold 100 mm Na 2 CO 3 and 4 ml of ethyl acetate followed by vortex mixing for extraction as described below. The final midazolam concentration ranged from 2.5 to 320 M, and the linearity of metabolism had been confirmed under the above conditions in terms of both protein concentration and incubation time. Data are presented as the means S.D. of triplicate experiments. Inhibition of Midazolam Metabolism by Macrolides. After 5 min of preincubation at 37 C of a reaction mixture (0.64 ml) consisting of human liver microsomes and an NADPH-generating system in potassium phosphate buffer containing EDTA as described above, 80 l of macrolide solution (10% acetone solution for erythromycin and clarithromycin; 20% acetone solution for azithromycin) was added and then preincubated further at 37 C for 0, 5, 10, or 20 min. Then, 80 l of midazolam in 20% acetone solution was added and incubated at 37 C for another 3 min. The enzyme reaction was terminated as described above. The final concentration of midazolam was set at 200 M, whereas that of erythromycin and clarithromycin ranged from 5 to 100 M, and that of azithromycin ranged from 50 to 1000 M. Data are presented as means S.E. of three determinations. Quantification of Midazolam Metabolites by HPLC. - and 4-OH Midazolam in the incubation mixture were determined by an HPLC-UV detection method. One hundred microliters of 2 g/ml etizolam (methanol solution) was added to the extraction mixture as an internal standard and centrifuged at 1500g for 10 min after vortex mixing. Three milliliters of supernatant was evaporated to dryness under a gentle stream of nitrogen. The residues were reconstituted with 300 l of HPLC mobile phase as described below, and 50 l was injected into the HPLC column. The HPLC system consisted of a model LC-10AD pump (Shimadzu Ltd., Kyoto, Japan), a model SIL-10A sample injector (Shimadzu), a model SPD-10A UV absorbance detector (Shimadzu) set at 220 nm, and a Mightysil RP-18 reversed-phase column ( mm inner diameter, Kanto Chemical Co., Tokyo, Japan). The mobile phase consisted of a 12/5/7 (v/v) mixture of 10 mm potassium phosphate buffer (ph 7.4), methanol, and acetonitrile delivered at 1.0 ml/min. All chromatograms were recorded using a model C-R4 Chromato-Integrator (Shimadzu), and - and 4-OH midazolam were quantified based on their peak areas. Kinetic Analysis of Midazolam Metabolism by Pooled Human Liver Microsomes. The kinetic parameters (V max and K m ) for midazolam - and 4-hydroxylation by human liver microsomes were determined by the nonlinear least-squares regression program MULTI (Yamaoka et al., 1981) according to the following equation: v V max S / K m S (1) where v, V max, and K m represent the metabolic rate, the maximum metabolic rate, and the Michaelis constant, respectively. The intrinsic clearance (CL int ) for each metabolic pathway was obtained as the ratio of V max and K m. Kinetic Analysis of Enzyme Inactivation by Macrolides. Kinetic parameters for enzyme inactivation were obtained as reported elsewhere (Ito et al., 1998). The logarithm of the remaining enzymatic activity (formation rate of or 4-OH midazolam) was plotted against the preincubation time, and the apparent inactivation rate constant (k obs ) was determined from the slope of the initial linear phase. Then, the value of k obs was plotted against the macrolide concentration ([I]), and the parameters (k inact, K app, and k d ) were obtained by the nonlinear least-squares regression method (MULTI) according to the following equation (Waley, 1985; Silverman, 1988): k obs k inact I / K app I k d (2) where k inact, K app, and k d represent the maximum inactivation rate constant, the apparent dissociation constant between the enzyme and the macrolides, and the spontaneous in vitro degradation rate constant of the enzyme in the absence of the inhibitor, respectively. Quantitative Prediction of the in Vivo Midazolam/Macrolide Interaction. The differential equations for active and inactive CYP3A in the liver (E act and E inact, respectively) can be described as follows: de act /dt k inact E act f b I liver /Kp / K app f b I liver /Kp k deg E 0 E act de inact /dt k inact E act f b I liver /Kp / K app f b I liver /Kp k deg E inact where k deg, Kp, f b, I liver, and E 0 represent the degradation rate constant (turnover rate constant) of CYP3A, liver-to-blood concentration ratio of macrolides, unbound fraction of macrolides in blood, macrolide concentration in the liver, and total concentration of CYP3A, respectively. The initial conditions (at t 0) are E act E 0 and E inact 0. In the absence of macrolides, the CYP3A content in the liver is at steady state and the degradation rate (k deg E 0 ) is equal to the synthesis rate, which was assumed to be unaffected by macrolides. It was also assumed in the above equations that the k deg for the inactive enzyme is the same as that for the active enzyme. The differential equations for midazolam (S) and macrolides (I) can be expressed as follows according to the perfusion model (Fig. 1): For midazolam: V liver d S liver /dt Q S pv Q S liver /Kp f b CL int S liver /Kp (5) CL int V max,1 / K m,1 f b S liver /Kp V max,2 / K m,2 f b S liver /Kp (6) For macrolides: (3) (4) V max,1 V max,1 0 E act,1 /E 0 (7) V max,2 V max,2 0 E act,2 /E 0 (8) V pv d S pv /dt Q S sys V abs Q S pv (9) V abs k a Dose F a F g e ka t (10) V sys d S sys /dt Q S liver /Kp Q S sys (11) V liver d I liver /dt Q I pv Q I liver /Kp f b CL int I liver /Kp (12) CL int V max / K m f b I liver /Kp (13) V pv d I pv /dt Q I sys V abs Q I pv (14) V abs k a Dose F a F g e ka t (15) V sys d I sys /dt Q I liver /Kp Q I sys CL r I sys (16) where V liver and V pv represent the volume of liver and portal vein, respectively; V sys represents the volume of distribution in the central compartment; S liver represents the concentration in the liver; S pv and I pv represent the concentration in the portal vein; S sys and I sys represent the concentration in the central compartment; Q represents the blood flow rate; CL r represents the renal clearance; CL int represents the intrinsic metabolic clearance; K m represents the Michaelis constant; K m,1 and K m,2 represent the K m for - and 4-hydroxylation of midazolam, respectively; V max represents the maximum rate of metabolism; V max,1 and V max,2 represent the V max for - and 4-hydroxylation of midazolam, respectively; E act,1 and E act,2 represent the E act for - and 4-hydroxylation of midazolam, respectively; V abs represents the absorption velocity; k a represents the first-order absorption rate constant; F a represents the fraction absorbed from the gastrointestinal tract; and F g represents the intestinal availability. In the case of intravenous administration of midazolam, the absorption term in eq. 9 (V abs ) was deleted, and the dose (nmol) was used as the initial value of V sys S sys.

3 PREDICTION OF MIDAZOLAM/MACROLIDES INTERACTION 947 FIG. 1.Physiological model for the description of the time-profiles of midazolam and macrolide concentrations. The following assumptions were made in the above mass-balance equations. 1. Midazolam is administered intravenously or orally and macrolides are administered orally. 2. Midazolam is eliminated only by the liver. 3. The distribution of midazolam and macrolides in the liver rapidly reaches equilibrium, and the unbound concentrations in the hepatic vein are equal to those in the liver at equilibrium (well stirred model). 4. Only the unbound molecule in the liver is subject to metabolism. 5. The contribution of CYP3A to the total elimination of macrolides in the liver is small (i.e., the elimination of a macrolide itself is not altered by inactivation of CYP3A). 6. Gastrointestinal absorption can be described by a first-order rate constant. The pharmacokinetic parameters of midazolam and macrolides were determined from data in the literature (Tables 1 and 2). Using the program STELLA II (High Performance Systems, Inc., Hanover, NH), and kinetic parameters for CYP3A inactivation obtained in in vitro studies, the above differential equations were numerically solved to simulate the time courses of the macrolide concentration in blood, the active CYP3A content in the liver (E act ), and midazolam concentration in blood. According to clinical reports, the dosing schedules were assumed as follows: in the case of intravenous administration of midazolam, erythromycin (500 mg 681 mol, t.i.d. for 6 days) followed by midazolam (9.67 mol) (Olkkola et al., 1993) or clarithromycin (500 mg 668 mol, b.i.d. for 7 days), followed by midazolam (9.82 mol) (Gorski et al., 1998). In the case of oral administration of midazolam, erythromycin (500 mg 681 mol, t.i.d. for 5 or 6 days) followed by midazolam (15 mg 46 mol) (Olkkola et al., 1993; Zimmermann et al., 1996), clarithromycin (250 TABLE 1 Pharmacokinetic parameters of midazolam and enzyme turnover used in the simulation Parameter Value Source Reference Midazolam Dose 46,045 (nmol) A, B, C 12,279 (nmol) D 9,669, 9,823 (nmol) A, D F a 1 assumed F g ( EM), ( EM) A 0.42 ( CAM), 0.83 ( CAM) D 0.42 ( AZM), 0.42 ( AZM) f b calculated from f u 0.022, RB 0.66 E, D Kp 1 assumed V liver 2,800 (ml) V pv 70 (ml) Q 1,610 (ml/min) CL int 15,000 (ml/min) CL int,1 14,130 (ml/min) -hydroxylation, 84.8% of CL int (Table 3) CL int,2 870 (ml/min) 4-hydroxylation, 15.2% of CL int (Table 3) K m,1 3.8 ( M) Table 3 V max,1 53,700 (nmol/min) calculated from K m,1 CL int,1 K m, ( M) Table 3 V max,2 11,700 (nmol/min) calculated from K m,2 CL int,2 k a 0.02 (min 1 ) V sys 50,000 (ml) Enzyme k deg (min 1 ) rat P450 F E 0 5 (nmol/g liver) CYP3A4 content in the liver G EM, erythromycin; CAM, clarithromycin; AZM, azithromycin; f u, unbound fraction in plasma; RB, blood to plasma concentration ratio. Calculated using eqs. 5 to 11 to fit the reported midazolam blood concentration (Olkkola et al., 1993). A, Olkkola et al., 1993; B, Zimmermann et al., 1996; C, Yeates et al., 1996; D, Gorski et al., 1998; E, Lown et al., 1995; F, Shiraki and Guengerich, 1984; G, Iwatsubo et al., 1997.

4 948 ITO ET AL. TABLE 2 Pharmacokinetic parameters of macrolides used in the simulation Parameter Value Schedule or Source Reference Erythromycin Dose 681,200 (nmol) t.i.d. A, B F a F g 0.58 Unpublished observation f b 0.16 H Kp 1 assumed V liver 2,800 (ml) V pv 70 (ml) Q 1,610 (ml/min) CL int 2,940 (ml/min) k a (min 1 ) CL r 0 (ml/min) V sys 18,000 (ml) T lag 96 (min) I Clarithromycin Dose 334,242, 668,485 (nmol) b.i.d. C, D F a F g 1 f b 0.28 calculated from f u 0.18, RB 0.64 J, H Kp 1 assumed V liver 2,800 (ml) V pv 70 (ml) Q 1,610 (ml/min) CL int 2,176 (ml/min) k a (min 1 ) CL r 100 (ml/min) V sys 100,000 (ml) T lag 12 (min) K Azithromycin Dose 636,943 (nmol) o.d. B, C F a F g 0.86 L f b 0.7 L Kp 1 assumed V liver 2,800 (ml) V pv 70 (ml) Q 1,610 (ml/min) CL int 1,800 (ml/min) k a (min 1 ) CL r 100 (ml/min) V sys 24,500 (ml) Calculated using eqs. 12 to 16 to fit the reported erythromycin (Birkett et al., 1990), clarithromycin (Chu et al., 1992), or azithromycin blood concentration (Foulds et al., 1990). A, Olkkola et al., 1993; B, Zimmermann et al., 1996; C, Yeates et al., 1996; D, Gorski et al., 1998; H, Benet et al., 1996; I, Birkett et al., 1990; J, Yago et al., 1996; K, Chu et al., 1992; L, Foulds et al., TABLE 3 Kinetic parameters for midazolam hydroxylation by human liver microsomes K m V max CL int M nmol/min/mg protein ml/min/mg protein % -Hydroxylation Hydroxylation TABLE 4 Kinetic parameters for inhibition of midazolam hydroxylation by macrolides Erythromycin Clarithromycin Azithromycin -OH 4-OH -OH 4-OH -OH 4-OH K app ( M) k inact (min 1 ) k d (min 1 ) mg 334 mol, b.i.d. for 5 days) followed by midazolam (15 mg 46 mol) (Yeates et al., 1996), clarithromycin (500 mg 668 mol, b.i.d. for 7 days) followed by midazolam (4 mg 12 mol) (Gorski et al., 1998), or azithromycin (500 mg 637 mol, o.d. for 3 days) followed by midazolam (15 mg 46 mol) (Yeates et al., 1996; Zimmermann et al., 1996). The AUCs from time 0 to infinity of the simulated midazolam concentration profiles were compared with the reported values. Results In Vitro Metabolism of Midazolam by Human Liver Microsomes. The metabolism of midazolam by human liver microsomes followed Michaelis-Menten kinetics with the kinetic constants summarized in Table 3. The K m value for 4-hydroxylation of midazolam was about 3.5-fold higher than that for the -hydroxylation pathway,

5 PREDICTION OF MIDAZOLAM/MACROLIDES INTERACTION 949 FIG. 2.Inhibitory effect of macrolides on midazolam hydroxylation by human liver microsomes. The y-axis represents the formation rate of -OH midazolam (A, C, and E) or 4-OH midazolam (B, D, and F). The control activities were A, 3.77; B, 1.63; C, 4.52; D, 1.94; E, 3.14; and F, 1.38 nmol/min/mg protein, respectively. whereas the V max was greater for -hydroxylation. Consequently, the CL int (V max /K m ratio) for the - and 4-hydroxylation pathways was 84.8% and 15.2%, respectively, of the total CL int, indicating that -hydroxylation is the major metabolic pathway of midazolam. This result is consistent with the previous finding by Gorski et al. (1994). Inhibition of Midazolam Metabolism by Macrolide Antibiotics. Figure 2 shows the effect of macrolide concentration and preincubation time on midazolam metabolism by human liver microsomes. Midazolam metabolism was not inhibited without preincubation, even if the macrolide concentration was increased. The degree of inhibition depended on the preincubation time and the macrolide concentration. -Hydroxylation of midazolam by human liver microsomes was reduced to 43.1%, 39.4%, and 67.5% of the control value following a 20-min preincubation in the presence of 100 M erythromycin, 100 M clarithromycin, and 1000 M azithromycin, respectively. Similar results were obtained for the 4-hydroxylation pathway. The calculated kinetic parameters for CYP3A inactivation are summarized in Table 4. The data points of the 0- to 10-min preincubation were considered to reflect the initial inactivation rate and were used to estimate the values of k obs. For each macrolide, the obtained values of both K app and k inact were almost identical for both hydroxylation pathways of midazolam. Quantitative Prediction of the Midazolam/Macrolide Interaction. Concentration profiles of midazolam and macrolides simulated by the kinetic parameters in Tables 1 and 2 were compared with the reported profiles (Birkett et al., 1990; Foulds et al., 1990; Chu et al., 1992; Olkkola et al., 1993). Figure 3 shows the concentration profiles of midazolam in blood after a single intravenous or oral administration and those of macrolides after a single oral administration. In all cases, the simulated and reported profiles were comparable. Furthermore, the simulated profiles were constant after each dose following repeated administration (Figs. 4 and 5), which was also consistent with the previous findings that no accumulation was observed for each macrolide following repeated administration (Smith et al., 1953; Suwa et al., 1988; Foulds et al., 1990). These findings indicate the validity of the pharmacokinetic parameters used in the present simulation. Figures 4 and 5 also show the simulated effects of erythromycin and clarithromycin, respectively, on the active CYP3A content in the liver and midazolam concentrations in blood. The results are also summarized in Table 5. Following administration of erythromycin (500 mg t.i.d. for 6 days), the active CYP3A was predicted to fall gradually, and a maximum 45% inactivation of CYP3A involved in both hydroxylation pathways of midazolam was predicted after 3 days of administration of erythromycin, with no further inactivation thereafter [Fig. 4A, (2)]. The concentration of midazolam in blood was predicted to increase following administration of erythromycin, and the predicted AUC increase compared with control group was 1.7- and 3.0-fold after intravenous and oral administration of midazolam, re- FIG. 3.Comparison between observed (closed circles) and simulated (dashed line) concentration profiles of midazolam and macrolides. A and B, midazolam concentration profiles in blood after single intravenous or oral administration of midazolam (3 or 15 mg) (Olkkola et al., 1993). C, erythromycin concentration profile in blood after single oral administration of erythromycin (250 mg) (Birkett et al., 1990). D, clarithromycin concentration profile in blood after single oral administration of clarithromycin (200 mg) (Chu et al., 1992). E, azithromycin concentration profile in blood after single oral administration of azithromycin (500 mg) (Foulds et al., 1990).

6 950 ITO ET AL. spectively [Fig. 4A, (3) and (4)]. In another case of erythromycin administration (500 mg t.i.d. for 5 days), a 2.9-fold increase was predicted in the AUC of midazolam after oral administration [Fig. 4B, (3)]. Following administration of clarithromycin (500 mg b.i.d. for 7 days), a maximum of 34 and 38% inactivation of the CYP3A concerned with the - and 4-hydroxylation pathway, respectively, of midazolam was predicted after 3 days of administration, with no further inactivation thereafter [Fig. 5A, (2)]. The predicted increase in midazolam AUC was 2.0- and 2.5-fold after intravenous and oral administration, respectively [Fig. 5A, (3) and (4)]. In another case of clarithromycin administration (250 mg b.i.d. for 5 days), a maximum of 20 and 23% inactivation of the CYP3A concerned with - and 4-hydroxylation pathway, respectively, of midazolam was predicted after 3 days of administration, with no further inactivation thereafter [Fig. 5B, (2)]. A 2.1-fold increase was predicted in the AUC of midazolam after oral administration [Fig. 5B, (3)]. In the case of azithromycin administration (500 mg o.d. for 3 days), only 1% of the CYP3A concerned with both hydroxylation pathways was predicted to be inactivated after 2 days of administration, with no further inactivation thereafter (data not shown). Almost no change was predicted in the AUC of midazolam after oral administration (Table 5). FIG. 4.Simulated profiles of erythromycin concentration in blood (1), active CYP3A content in the liver (2), and midazolam concentration in blood [(3) and (4)]. In (2), (3), and (4), solid line represents control and dashed line represents erythromycin.

7 PREDICTION OF MIDAZOLAM/MACROLIDES INTERACTION 951 FIG. 5.Simulated profiles of clarithromycin concentration in blood (1), active CYP3A content in the liver (2), and midazolam concentration in blood [(3) and (4)]. In (2), (3), and (4), solid line represents control and dashed line represents clarithromycin.

8 952 ITO ET AL. Pretreatment TABLE 5 Prediction of the increase in midazolam AUC by macrolides Midazolam Predicted AUC Ratio k deg (min 1 ) AUC Ratio (mean S.D.) Reported Reference Erythromycin 500 mg t.i.d. for 6 days i.v Olkkola et al., mg t.i.d. for 6 days p.o Olkkola et al., mg t.i.d. for 5 days p.o Zimmermann et al., 1996 Clarithromycin 500 mg b.i.d. for 7 days i.v Gorski et al., mg b.i.d. for 7 days p.o Gorski et al., mg b.i.d. for 5 days p.o Yeates et al., 1996 Azithromycin 500 mg o.d. for 3 days p.o Olkkola et al., mg o.d. for 3 days p.o Gorski et al., 1998 Discussion Various types of macrolide antibiotics are now being used in clinical practice. It has been demonstrated that macrolides inhibit the metabolic enzyme CYP3A by forming an inactive complex in vitro (Periti et al., 1992). They are also reported to cause many in vivo interactions with drugs that are substrates of CYP3A, and the severity of such interactions varies from macrolide to macrolide (Periti et al., 1992; Westphal, 2000). Whereas erythromycin and clarithromycin markedly affect the AUC of midazolam, a CYP3A substrate, in humans, the effect of azithromycin is reported to be small (Backman et al., 1994; Yeates et al., 1996; Zimmermann et al., 1996). Periti et al. (1992) have classified troleandomycin and erythromycin, 14-carbon member ring macrolides, as belonging to the first group with a high drug interaction potential; clarithromycin and roxithromycin (14-carbon), midecamycin and josamycin (16-carbon) belong to the second group, with a low interaction potential; and azithromycin (15-carbon), rokitamycin, and spiramycin (16-carbon) represent the third group with no interaction potential. The inhibitory effects of macrolides on CYP3A have been analyzed based on competitive or noncompetitive inhibition (Echizen et al., 1993; Wrighton and Ring, 1994; von Moltke et al., 1996; Thummel and Wilkinson, 1998). On the other hand, complex formation with P450 is also reported to be involved in the inhibition by macrolides (Murray, 1987; Periti et al., 1992). Tinel et al. (1989) evaluated the complex formation potential of macrolides using the liver microsomes from dexamethasone-treated rats and reported that the rate of complex formation was highest for troleandomycin, followed by erythromycin, and that the rates for clarithromycin and roxithromycin were lower than those for erythromycin. Using human liver microsomes, Yamazaki and Shimada (1998) have reported similar results, showing that the complex formation potential is greatest for troleandomycin, followed by erythromycin, with roxithromycin having the lowest potential. In the prediction of in vivo drug interactions involving this type of enzyme inhibition from in vitro studies, the exposure time of the enzyme to the inhibitor should be taken into account as well as the turnover rate of the enzyme. We have already succeeded in quantitatively predicting the triazolam/erythromycin interaction in humans based on a physiologically based pharmacokinetic model taking the type of inhibition into consideration (Kanamitsu et al., 2000b). Yamano et al. (2001) have also presented a successful prediction of interaction between midazolam and erythromycin using a similar model. In the present study, a similar model was applied to the interaction between midazolam and macrolides with different inhibitory potentials. At first, we tried to predict the increase in midazolam AUC assuming competitive inhibition of CYP3A by erythromycin and clarithromycin (Table 6). Gascon and Dayer (1991) have reported the inhibition constant (K i ) of erythromycin on midazolam -hydroxylation by human liver microsomes, assuming competitive inhibition. The maximum unbound concentration of erythromycin at the inlet to the liver (I in,u ) was estimated by I in,u (I max k a Dose F a /Q) f b, where I max is the maximum concentration in the systemic circulation (Ito et al., 1998). Since midazolam is eliminated from human body predominantly via CYP3A-mediated metabolism (Smith et al., 1981), the AUC increase by erythromycin was predicted by 1 I in,u /K i (Ito et al., 1998). The increase in the midazolam AUC produced by clarithromycin was also predicted in the same way, except that the IC 50 value reported by Gascon and Dayer (1991) was used instead of the K i, which was not available for clarithromycin. As shown in Table 6, almost no increase was predicted in the AUC of midazolam, indicating that the reported 3.6- to 7.0-fold increase in vivo (Olkkola et al., 1993; Yeates et al., 1996; Zimmermann et al., 1996; Gorski et al., 1998) cannot be explained by competitive inhibition of the enzyme. In the present in vitro studies using human liver microsomes, midazolam metabolism was not inhibited without preincubation, even if the concentration of the macrolides was increased, and the degree of inhibition depended on the preincubation time and the macrolide concentration (Fig. 2). These findings indicate that the inhibitory effect of macrolides on midazolam metabolism is predominantly TABLE 6 Prediction of the increase in midazolam AUC assuming a competitive inhibition of CYP3A by erythromycin or clarithromycin I max k a DF a /Q I in,u a K i b Predicted AUC Ratio c M M M M Erythromycin (500 mg p.o.) Clarithromycin (250 mg p.o.) Clarithromycin (500 mg p.o.) a I in,u (I max k a DF a /Q) f u, where I max is maximum systemic concentration, k a is absorption rate constant, D is dose, F a is fraction absorbed, Q is hepatic blood flow, and f u is unbound fraction. b K i or IC 50 for the inhibition of midazolam -hydroxylation by human liver microsomes (Gascon and Dayer, 1991). c Predicted AUC ratio 1 I in,u /K i.

9 PREDICTION OF MIDAZOLAM/MACROLIDES INTERACTION 953 caused by mechanism-based inhibition of CYP3A, with little contribution from competitive inhibition. The estimated values of K app and k inact for each macrolide were almost the same for both midazolam hydroxylation pathways (Table 4), and the values for erythromycin were comparable to those for triazolam metabolism reported by Kanamitsu et al. (2000b). As shown in Table 5, using the k deg of min 1, the increase in midazolam AUC was slightly underestimated. The ratio of the predicted and reported increase in midazolam AUC was between 0.6 and 0.8 except for the case of oral administration of midazolam after clarithromycin treatment (500 mg b.i.d. for 7 days) (Gorski et al., 1998), in which case the in vivo interaction was significantly underestimated. One of the reasons for this may be incorrect estimation of the observed AUC due to the lack of midazolam concentration data in the elimination phase. The average turnover rate constant (k deg ) of rat P450 ( min 1 ) (Shiraki and Guengerich, 1984) was used in the present simulation because the corresponding value for human CYP3A has not been reported. When the minimum reported value of k deg ( min 1 ) was used, almost the same results were obtained for clarithromycin and azithromycin, whereas the predicted degree of in vivo interaction was increased in the case of erythromycin, compared with using the k deg of min 1 (Table 5). Once an enzyme is inactivated in vivo by a mechanism-based inhibitor such as macrolides, the recovery of the metabolic activity depends solely on the synthesis of the enzyme. Thus, the turnover rate of the enzyme is one of the most important parameters in the prediction of interactions involving mechanism-based inhibition, and in the present study, this was found to affect the results in some cases. In cases where the turnover rate of the human enzyme is unavailable, it seems to be important to alter this parameter to some extent in the simulation, referring to animal data, to predict the range of the interaction. In addition, the liver-to-blood concentration ratio (Kp) of midazolam and macrolides was assumed to be 1 in the present prediction because this cannot be measured in humans. To examine the effect of Kp on the prediction, the Kp of midazolam was changed to 0.1 and 10, whereas that of the macrolides was fixed at 1, and the same simulation was conducted using the values of V d and CL int of midazolam redetermined to fit its concentration profile in blood. In another case, the Kp of macrolides was changed to 10, with that of midazolam being fixed at 1 and the simulation was conducted in a similar manner. As a result, the predicted increase in midazolam AUC was 2.8- to 3.0-fold for erythromycin (500 mg t.i.d. for 6 days), 2.8- to 2.9-fold for erythromycin (500 mg t.i.d. for 5 days), 2.3- to 2.5-fold for clarithromycin (500 mg b.i.d. for 7 days), 2.1-fold for clarithromycin (250 mg b.i.d. for 5 days), and 1.0- to 1.1-fold for azithromycin (500 mg o.d. for 3 days), suggesting that the value of Kp has little impact on the prediction. Although only the interaction involving the hepatic enzyme has been evaluated in the present in vitro study, it has been reported that CYP3A is also present in the small intestine (Paine et al., 1997) and that midazolam is metabolized by human intestinal microsomes in vitro (Thummel et al., 1996). In vivo human studies have revealed that the hepatic and intestinal availability of midazolam is 0.74 and 0.42, respectively (Gorski et al., 1998), demonstrating that there is also significant intestinal metabolism of midazolam in vivo. Furthermore, the intestinal availability (F g ) of midazolam is reported to increase approximately 2-fold by pretreatment with either erythromycin or clarithromycin (Olkkola et al., 1993; Gorski et al., 1998), indicating an interaction involving intestinal first-pass metabolism of midazolam. In the present study, the reported values of midazolam F g, with and without pretreatment with macrolides, were used in the prediction. However, in the case of interaction with azithromycin, the control F g value was used as a constant value assuming the absence of an interaction involving intestinal metabolism, because of lack of any reported value of F g after pretreatment with azithromycin. The effects of 10 min of preincubation with erythromycin (100 M), clarithromycin (100 M), and azithromycin (1000 M) on midazolam metabolism were similar between liver and jejunal microsomes from the same three donors, the degree of inhibition by azithromycin being much smaller than that by erythromycin or clarithromycin (unpublished observation). This finding indicates the validity of the above assumption that azithromycin does not cause an interaction in the small intestine. However, because the F g values used were in vivo data ignoring the time course of intestinal metabolism, efforts are being made by our group to predict the interaction involving the small intestine from in vitro data. Recently, Mayhew et al. (2000) proposed a more simple method of predicting in vivo drug interactions involving metabolic intermediate complex formation. Based on the inhibitor-induced change in the steady-state enzyme level ([E]ss), the degree of increase in the AUC after oral administration can be estimated by the following equation: AUC/AUC( I) CL int ( I)/CL int E SS ( I)/[E] SS k deg / k deg k inact I]/(K app I (17) TABLE 7 Prediction of the increase in midazolam AUC by macrolides using the steady-state enzyme level Predicted AUC Ratio a Reported Pretreatment [I] I max b I max,u c I in,u d k deg (min 1 ) AUC Ratio (mean S.D.) Reference Erythromycin 500 mg t.i.d. for 6 days Olkkola et al., mg t.i.d. for 5 days Zimmermann et al., 1996 Clarithromycin 500 mg b.i.d. for 7 days Gorski et al., mg b.i.d. for 5 days Yeates et al., 1996 Azithromycin 500 mg o.d. for 3 days Zimmermann et al., mg o.d. for 3 days Yeates et al., 1996 a Predicted AUC ratio (k deg k inact [I]/(K app [I]))/k deg using K app and k inact for the inhibition of midazolam -hydroxylation by human liver microsomes. I max,i max,u or I in,u for [I]. b I max, maximum concentration of inhibitor in the systemic blood. c I max,u, maximum unbound concentration of inhibitor in the systemic blood. I max,u I max f u where f u is unbound fraction. d I in,u, unbound concentration of inhibitor at the inlet to the liver. I in,u (I max k a DF a /Q) f u where k a is absorption rate constant, D is dose, F a is fraction absorbed, and Q is hepatic blood flow.

10 954 ITO ET AL. where ( I) represents the value in the presence of the inhibitor and [I] represents the concentration of inhibitor. We have tried to predict the above-described interactions involving macrolides according to eq. 17, using the k deg value for rat P450 as above and the K app and k inact values for -hydroxylation of midazolam obtained in the present study (Table 7). The maximum concentration in blood (I max ), maximum unbound concentration in blood (I max,u ), or maximum unbound concentration at the inlet to the liver (I in,u ) was used as [I]. As shown in Table 7, the predicted increase in the midazolam AUC was comparable to the reported value when I in,u was used. Applying this methodology to other interactions involving mechanism-based inhibitors should provide information for determining what concentration of inhibitor (I max,i in,u etc.) is appropriate for [I] in eq. 17. The degree of interaction with CYP3A substrates varies among the macrolides. In the present study, some of these interactions have been successfully predicted from in vitro data. If an interaction involving mechanism-based inhibition of the enzyme is analyzed assuming a competitive inhibition, the in vivo interaction should be greatly underestimated. Quantitative predictions from in vitro data taking the inhibition type into consideration are essential for avoiding toxic interactions in clinical practice. Acknowledgments. We express our appreciation to Dr. Yuichi Sugiyama (Graduate School of Pharmaceutical Sciences, University of Tokyo) for helpful discussions and reviewing the manuscript. We are grateful to Dr. Noriaki Shimada (Daiichi Pure Chemicals Co., Ltd.) for kindly providing human liver microsomes from BD Gentest. References Backman JT, Aranko K, Himberg JJ, and Olkkola KT (1994) A pharmacokinetic interaction between roxithromycin and midazolam. Eur J Clin Pharmacol 46: Benet LZ, Die S, and Schwartz JB (1996) Design and optimization of dosage regimens; pharmacokinetic data, in Goodman and Gilman s The Pharmacological Basis of Therapeutics, 9th ed. (Hardman JG and Limbird LE, eds) pp , McGraw-Hill, New York. Birkett DJ, Robson RA, Grgurinovich N, and Tonkin A (1990) Single oral dose pharmacokinetics of erythromycin and roxithromycin and the effects of chronic dosing. Ther Drug Monit 12: Chu SY, Sennello LT, Bunnell ST, Varga LL, Wilson DS, and Sonders RC (1992) Pharmacokinetics of clarithromycin, a new macrolide, after single ascending oral doses. Antimicrob Agents Chemother 36: Echizen H, Kawasaki H, Chiba K, Tani M, and Ishizaki T (1993) A potent inhibitory effect of erythromycin and other macrolide antibiotics on the mono-n-dealkylation metabolism of disopyramide with human liver microsomes. J Pharmacol Exp Ther 264: Food and Drug Administration (1990) Transcript of Proceedings. Pulmonary-Allergy Drugs Advisory Committee, June 11, Department of Health and Human Services, Public Health Service, Rockville, MD. Foulds G, Shepard RM, and Johnson RB (1990) The pharmacokinetics of azithromycin in human serum and tissues. J Antimicrob Chemother 25 (Suppl): Gascon MP and Dayer P (1991) In vitro forecasting of drugs which may interfere with the biotransformation of midazolam. Eur J Clin Pharmacol 41: Gorski JC, Hall SD, Jones DR, VandenBranden M, and Wrighton SA (1994) Regioselective biotransformation of midazolam by members of the human cytochrome P450 3A (CYP3A) subfamily. Biochem Pharmacol 47: Gorski JC, Jones DR, Haehner-Daniels BD, Hamman MA, O Mara EM Jr, and Hall SD (1998) The contribution of intestinal and hepatic CYP3A to the interaction between midazolam and clarithromycin. Clin Pharmacol Ther 64: Honig PK, Wortham DC, Zamani K, and Cantilena LR (1994) Comparison of the effect of the macrolide antibiotics erythromycin, clarithromycin and azithromycin on terfenadine steadystate pharmacokinetics and electrocardiographic parameters. Drug Investig 7: Ito K, Iwatsubo T, Kanamitsu S, Ueda K, Suzuki H, and Sugiyama Y (1998) Prediction of pharmacokinetic alterations caused by drug-drug interactions: metabolic interaction in the liver. Pharmacol Rev 50: Iwatsubo T, Hirota N, Ooie T, Suzuki H, Shimada N, Chiba K, Ishizaki T, Green CE, Tyson CA, and Sugiyama Y (1997) Prediction of in vivo drug metabolism in the human liver from in vitro metabolism data. Pharmacol Ther 73: Jurima-Romet M, Crawford K, Cyr T, and Inaba T (1994) Terfenadine metabolism in human liver. In vitro inhibition by macrolide antibiotics and azole antifungals. Drug Metab Dispos 22: Kanamitsu S, Ito K, Green CE, Tyson CA, Shimada N, and Sugiyama Y (2000b) Prediction of in vivo interaction between triazolam and erythromycin based on in vitro studies using human liver microsomes and recombinant human CYP3A4. Pharm Res (NY) 17: Kanamitsu S, Ito K, Okuda H, Ogura K, Watabe T, Muro K, and Sugiyama Y (2000a) Prediction of in vivo drug-drug interactions based on mechanism-based inhibition from in vitro data: inhibition of 5-fluorouracil metabolism by (E)-5-(2-Bromovinyl)uracil. Drug Metab Dispos 28: Lown KS, Thummel KE, Benedict PE, Shen DD, Turgeon DK, Berent S, and Watkins PB (1995) The erythromycin breath test predicts the clearance of midazolam. Clin Pharmacol Ther 57: Mayhew BS, Jones DR, and Hall SD (2000) An in vitro model for predicting in vivo inhibition of cytochrome P450 3A4 by metabolic intermediate complex formation. Drug Metab Dispos 28: Murray M (1987) Mechanisms of the inhibition of cytochrome P-450-mediated drug oxidation by therapeutic agents. Drug Metab Rev 18: Olkkola KT, Aranko K, Luurila H, Hiller A, Saarnivaara L, Himberg JJ, and Neuvonen PJ (1993) A potentially hazardous interaction between erythromycin and midazolam. Clin Pharmacol Ther 53: Paine MF, Khalighi M, Fisher JM, Shen DD, Kunze KL, Marsh CL, Perkins JD, and Thummel KE (1997) Characterization of interintestinal and intraintestinal variations in human CYP3Adependent metabolism. J Pharmacol Exp Ther 283: Periti P, Mazzei T, Mini E, and Novelli A (1992) Pharmacokinetic drug interactions of macrolides. Clin Pharmacokinet 23: Shiraki H and Guengerich FP (1984) Turnover of membrane proteins: kinetics of induction and degradation of seven forms of rat liver microsomal cytochrome P-450, NADPH-cytochrome P-450 reductase and epoxide hydrolase. Arch Biochem Biophys 235: Silverman RB (1988) Mechanism-based enzyme inactivation, in Chemistry and Enzymology, vol 1, pp 3 30, CRC Press LLC, Boca Raton, FL. Smith JW, Dyke RW, and Griffith RS (1953) Absorption following oral administration of erythromycin. J Am Med Assoc 151: Smith MT, Eadie MJ, and Brophy TO (1981) The pharmacokinetics of midazolam in man. Eur J Clin Pharmacol 19: Suwa T, Urano H, Kodama T, Nakamura M, and Watanabe T (1988) Metabolic fate of TE-031 (A-56268) (8) absorption and excretion in humans (bioassay). Chemotherapy 36: Thummel KE, O Shea D, Paine MF, Shen DD, Kunze KL, Perkins JD, and Wilkinson GR (1996) Oral first-pass elimination of midazolam involves both gastrointestinal and hepatic CYP3Amediated metabolism. Clin Pharmacol Ther 59: Thummel KE and Wilkinson GR (1998) In vitro and in vivo drug interactions involving human CYP3A. Annu Rev Pharmacol Toxicol 38: Tinel M, Descatoire V, Larrey D, Loeper J, Labbe G, Letteron P, and Pessayre D (1989) Effects of clarithromycin on cytochrome P-450. Comparison with other macrolides. J Pharmacol Exp Ther 250: von Moltke LL, Greenblatt DJ, Harmatz JS, Duan SX, Harrel LM, Cotreau-Bibbo MM, Pritchard GA, Wright CE, and Shader RI (1996) Triazolam biotransformation by human liver microsomes in vitro: effect of metabolic inhibitors and clinical confirmation of a predicted interaction with ketoconazole. J Pharmacol Exp Ther 276: Waley SG (1985) Kinetics of suicide substrates. Practical procedures for determining parameters. Biochem J 227: Westphal JF (2000) Macrolide-induced clinically relevant drug interactions with cytochrome P-450A (CYP) 3A4: an update focused on clarithromycin, azithromycin and dirithromycin. Br J Clin Pharmacol 50: Wrighton SA and Ring BJ (1994) Inhibition of human CYP3A catalyzed 1 -hydroxy midazolam formation by ketoconazole, nifedipine, erythromycin, cimetidine and nizatidine. Pharm Res (NY) 11: Yago K, Kuroyama M, Motohashi S, and Kumano K (1996) Protein binding of clarithromycin in patients with chronic renal failure (in Japanese). Jpn J Antibiot 49: Yamano K, Yamamoto K, Katashima M, Kotaki H, Takedomi S, Matsuo H, Ohtani H, Sawada Y, and Iga T (2001) Prediction of midazolam CYP3A inhibitors interaction in the human liver from in vivo/in vitro absorption, distribution and metabolism data. Drug Metab Dispos 29: Yamaoka K, Tanigawara Y, Nakagawa T, and Uno T (1981) A pharmacokinetic analysis program (MULTI) for microcomputer. J Pharmacobio-Dyn 4: Yamazaki H and Shimada T (1998) Comparative studies of in vitro inhibition of cytochrome P450 3A4-dependent testosterone 6beta-hydroxylation by roxithromycin and its metabolites, troleandomycin and erythromycin. Drug Metab Dispos 26: Yeates RA, Laufen H, and Zimmermann T (1996) Interaction between midazolam and clarithromycin: comparison with azithromycin. Int J Clin Pharm Ther 34: Zimmermann T, Yeates RA, Laufen H, Scharpf F, Leitold M, and Wildfeuer A (1996) Influence of the antibiotics erythromycin and azithromycin on the pharmacokinetics and pharmacodynamics of midazolam. Arzneim-Forsch 46: