QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Data Supplement All Versions of this Article: jpet.105.093229v1 316/1/336    most recent Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Obach, R. S. Articles by Tremaine, L. M. Search for Related Content PubMed PubMed Citation Articles by Obach, R. S. Articles by Tremaine, L. M.

ABSORPTION, DISTRIBUTION, METABOLISM, AND EXCRETION

The Utility of in Vitro Cytochrome P450 Inhibition Data in the Prediction of Drug-Drug Interactions

Pharmacokinetics, Dynamics, and Drug Metabolism (R.S.O., R.L.W., E.A.G., L.M.T.) and Clinical Pharmacokinetics and Pharmacodynamics (K.V.), Pfizer, Inc., Groton, Connecticut; and School of Pharmacy and Pharmaceutical Sciences, University of Manchester, Manchester, United Kingdom (J.B.H.)

Received July 25, 2005; accepted September 27, 2005.

	Abstract

Top Abstract Materials and Methods Results Discussion Conclusions References

 
The accuracy of in vitro inhibition parameters in scaling to in vivo drug-drug interactions (DDI) was examined for over 40 drugs using seven human P450-selective marker activities in pooled human liver microsomes. These data were combined with other parameters (systemic C_max, estimated hepatic inlet C_max, fraction unbound, and fraction of the probe drug cleared by the inhibited enzyme) to predict increases in exposure to probe drugs, and the predictions were compared with in vivo DDI gathered from clinical studies reported in the scientific literature. For drugs that had been tested as precipitants of drug interactions for more than one P450 in vivo, the order of inhibitory potencies in vitro generally aligned with the magnitude of the in vivo interactions. With the exception of many drugs known to be mechanism-based inactivators, the use of in vitro IC₅₀, the fraction of the affected drug metabolized by the target enzyme [f_m(CYP)] and an estimate of free hepatic inlet C_max, was generally successful in identifying those drugs that cause at least a 2-fold increase in the exposure to P450 marker substrate drugs. For CYP3A, incorporation of inhibition of both hepatic and intestinal metabolism was needed for the prediction of DDI. Many CYP3A inhibitors showed a different inhibitory potency for three different CYP3A marker activities; however, these differences generally did not alter the conclusions regarding whether a drug would cause a CYP3A DDI in vivo. Overall, these findings support the conclusion that P450 in vitro inhibition data are valuable in designing clinical DDI study strategies and can be used to predict the magnitudes of DDI.

In vitro drug-drug interaction data are necessary for devising mechanistically based clinical drug-drug interaction study strategies. The effects of new drugs on well characterized drug metabolism reactions known to be specific for various human drug-metabolizing enzymes are routinely examined using in vitro approaches. Frequently, human liver microsomes, a rich source of human drug-metabolizing enzymes such as cytochrome P450s, are used as an in vitro system. Of the human P450 enzymes, five have been described to contribute to the metabolism of the vast majority of drugs compared with other enzymes. These are CYP1A2, CYP2C9, CYP2C19, CYP2D6, and CYP3A4, albeit some other human P450 enzymes, such as CYP3A5, CYP2B6, and CYP2C8, have recently been gaining increased attention as potentially important drug-metabolizing enzymes.

To utilize in vitro data to develop strategies for in vivo drug interaction studies, a great deal of confidence must be placed in the capability to predict in vivo outcomes from in vitro information. Numerous efforts have been made to predict the magnitude of in vivo interactions from in vitro inhibition data, with varying degrees of success and failure. In principle, this should be attainable, because the process of predicting the magnitude in vivo drug-drug interactions from in vitro data is founded in well established pharmacokinetic theory. Relating in vitro potency for inhibition of drug-metabolizing enzymes to in vivo concentrations of the inhibitor, along with other information (e.g., plasma protein binding, dose, etc.), should yield projections of the magnitudes of drug-drug interactions (Yao and Levy, 2002; Houston and Galetin, 2003; Neal et al., 2003; Venkatakrishnan et al., 2003; Blanchard et al., 2004; Shou, 2005). However, in practice this has not necessarily been the case. Whether our lack of ability to do this for all drugs has been due to shortcomings in the gathering of accurate in vitro data or our lack of a complete understanding of the impact of in vivo phenomena (e.g., plasma protein binding, transporter mediated hepatic uptake, impact of intestinal metabolism, etc.) on drug-drug interactions remains an active area of research (Benet et al., 2003; Chien et al., 2003; Houston and Galetin, 2003; Venkatakrishnan et al., 2003; Ito et al., 2004).

However, even if our present knowledge were not to allow accurate prediction of the magnitude of drug-drug interactions from in vitro inhibition data in all cases, at the very least, these in vitro data should be able to indicate which enzymes will be more affected than others. Although phenomena such as active uptake into hepatocytes or the relevance of plasma protein binding may confound the ability to get an accurate prediction of the magnitude of a drug-drug interaction, such phenomena should confound this prediction for all of the hepatic drug-metabolizing enzymes equally. Therefore, the rank order of in vitro inhibition potencies and magnitudes of in vivo drug-drug interactions should be the same. If this premise is accepted, then by combining in vitro inhibition data with in vivo data targeting the most potently inhibited enzyme, the number of clinical drug-drug interaction studies needed to understand the interaction potential for a new drug should be able to be reduced. Such a strategy has been described (Huang et al., 1999). Regardless of whether quantitative prediction of the magnitude of drug interactions in vivo is possible, such an approach still allows the mechanistically based in vitro data to reduce the number of in vivo studies required.

The primary objectives of this work are to determine whether in vitro inhibition data can be used to quantitatively predict the magnitudes of DDI in vivo and to identify which in vivo parameters are needed for accurate prediction. Although such exercises have been described for various individual drugs and enzymes, in the present report, this was attempted across a wide range of drugs and for five different P450 enzymes. Additionally, because it has been suggested that measurement of more than one substrate in vitro activity may be needed to identify CYP3A inhibitors (Kenworthy et al., 1999; Bjornsson et al., 2003), we sought to determine whether inhibition data from one of these in vitro activity assays is superior to the others.

	Materials and Methods

Top Abstract Materials and Methods Results Discussion Conclusions References

 
Literature in Vivo Drug Interaction Data. Most of the drugs that were selected for study had been tested in vivo in humans with selective probe drugs for at least three of the following five enzymes: CYP1A2, CYP2C9, CYP2C19, CYP2D6, or CYP3A. These drugs are well absorbed and have not been shown to be profoundly affected by drug transporters; i.e., Biopharmaceutics Drug Disposition Class 1 and 2 drugs (Wu and Benet, 2005). The probes (object drugs) were theophylline or clozapine (CYP1A2), tolbutamide or warfarin (CYP2C9), mephenytoin or omeprazole (CYP2C19), desipramine, metoprolol, or dextromethorphan (CYP2D6), or midazolam, alprazolam, triazolam, buspirone, nifedipine, or simvastatin (CYP3A). In cases in which the same object/precipitant pair was tested in more than one study, the study showing the greatest increase was used to apply the most conservative analysis. Studies in which systemic AUC measurements were made to define the extent of the interaction were preferred followed by those reporting changes in other systemic exposure parameters (e.g., C_max, C_min, and t_1/2). If the only data available were urinary activity indices (e.g., dextromethorphan/dextrorphan or S-/R-mephenytoin metabolic ratios, or 4'-hydroxymephenytoin urinary recovery), these were used. Whenever available, drug interaction data from simultaneous oral administration were preferentially utilized versus other modes of administration. An extensive literature search yielded 21 precipitant drugs that met these criteria: cimetidine, citalopram, clarithromycin, diltiazem, disulfiram, erythromycin, fluconazole, fluoxetine, fluvoxamine, ketoconazole, metronidazole, nefazodone, paroxetine, propranolol, ranitidine, risperidone, roxithromycin, sertraline, terbinafine, ticlopidine, and troleandomycin.

Data for an additional set of 23 drugs that had been tested as potential inhibitors of CYP3A in clinical studies were also collected. These, combined with the aforementioned drugs, were used to assess the quantitative prediction of drug interactions occurring at CYP3A4. Finally, to increase the number of drugs that cause an interaction in vivo at CYP1A2, CYP2C9, CYP2C19, and CYP2D6, in vivo interaction data were gathered for enoxacin (CYP1A2), dicumarol and sulfaphenazole (CYP2C9), moclobemide (CYP2C19), and quinidine and diphenhydramine (CYP2D6).

In Vitro Data. Because in vitro inhibition constants can differ when measured in different laboratories, depending on various experimental details, these data were collected experimentally for the precipitant drugs using assays for five human P450s. The inhibition data were gathered using previously described validated methods (Walsky and Obach, 2004) for CYP1A2 (phenacetin O-deethylase), CYP2C9 (diclofenac 4'-hydroxylase), CYP2C19 (S-mephenytoin 4'-hydroxylase), CYP2D6 (dextromethorphan O-demethylase), and CYP3A (testosterone 6-hydroxylase, midazolam 1'-hydroxylase, and felodipine dehydrogenase). Human liver microsomes pooled from 53 individual donors were used as the source of enzyme activity. Protein concentrations employed were at or below 0.03 mg/ml (with the exception of CYP2C19, which was run at 0.2 mg/ml) to keep nonspecific binding to microsomes to a minimum. IC₅₀ values were measured using the substrates at a concentration equal to previously determined K_M values (Walsky and Obach, 2004). Inhibitors were examined up to a maximum concentration of 300 µM. IC₅₀ curves were fit using SigmaPlot (version 8; SPSS Inc., Chicago, IL). Because inhibition experiments were conducted at a substrate concentration equal to K_M in all cases, IC₅₀ values were converted to K_i values by dividing by two and assuming competitive inhibition.

Predictions of Drug Interactions for CYP1A2, CYP2C9, CYP2C19, and CYP2D6. For competitive inhibitors, the magnitude of the increase in exposure is related to the inhibitory potency (K_i), the concentration of inhibitor in vivo ([I]_{in vivo}), and the fraction of the clearance of the affected drug that occurs via metabolism by the inhibited P450 enzyme [f_m(CYP)]. This relationship has been described previously (Rowland and Matin, 1973; Ito et al., 2005).

(1)

Values for f_m(CYP) for the standard in vivo probes for CYP enzymes were estimated as follows. For CYP2C9, CYP2C19, and CYP2D6, studies of exposure differences for probe substrates between extensive and poor metabolizers can provide a reasonable estimate of fraction metabolized (Lennard et al., 1983; Kupfer et al., 1984; Brosen et al., 1993; Kirchheiner et al., 2002; Scordo et al., 2002; Yasui-Furukori et al., 2004). Values used for f_m(CYP) for CYP2C9 were 0.78, 0.91, and 0.45 for tolbutamide, S-warfarin, and rac-warfarin, respectively; those used for CYP2C19 were unity and 0.87 for S-mephenytoin and omeprazole, respectively; and those used for CYP2D6 were 0.9, 0.8, and unity for desipramine, metoprolol, and dextromethorphan/dextrorphan urinary ratio, respectively. For theophylline and clozapine, estimates of the contribution of CYP1A2 to total clearance (0.8 and 0.6, respectively) were made using a combination of quantitative human metabolism data reported from subjects receiving radiolabeled drugs (Monks et al., 1979; Dain et al., 1997) and in vitro data describing P450 enzymes involved in metabolic pathways.

Four different values for [I]_{in vivo} were tested for each inhibitor to determine which was the best value to use. These were the total systemic C_max, free systemic C_max, total hepatic inlet C_max estimated after oral administration, and free hepatic inlet C_max. Estimated maximum hepatic inlet concentrations for the inhibitors were calculated according to Kanamitsu et al. (2000),

(2)

in which F_a is the fraction of the inhibitor passing through the intestine unchanged (product of fraction absorbed and fraction un-metabolized by the intestine), k_a is the absorption rate constant, D is dose, and Q_h is human hepatic blood flow (21 ml/min/kg). Values for F_a for each of the inhibitors were estimated from literature values of oral bioavailability and hepatic extraction, with the latter value derived from intravenous clearance and hepatic blood flow. In the absence of such data, a value of unity was assumed as the most cautious possibility. A value of 0.03 min^-1 was used for k_a for all drugs.

Predictions of Drug Interactions for CYP3A. For the other CYP enzymes described in the preceding section, only the hepatic contribution was considered; however, for CYP3A, a contribution from gut was also considered. The relative contributions of intestinal and hepatic CYP3A to the extraction of the probe substrates were as follows: midazolam F_g = 0.57, F_h = 0.56; triazolam F_g = 0.64, F_h = 0.25; alprazolam F_g = 0.99, F_h = 0.89; buspirone F_g = 0.21, F_h = 0.24; simvastatin F_g = 0.66, F_h = 0.67; and nifedipine F_g = 0.74, F_h = 0.55. These estimates were made from reports on pharmacokinetics after administration to anhepatic patients or combination of clearance from intravenous administration and oral bioavailability or from estimates of hepatic and intestinal intrinsic clearance. Values used for clearance (CL_iv) for midazolam, triazolam, alprazolam, nifedipine, simvastatin, and buspirone were 6.6, 5.6, 0.74, 7, 7, and 19 ml/min/kg, respectively, along with f_u values of 0.02, 0.10, 0.29, 0.04, 0.06, and 0.05, respectively. The amount of CYP3A in the intestine has been reported to be approximately 1% of that in the liver (Thummel et al., 1997; Yang et al., 2004). For all of the CYP3A probe substrates, a value for f_m(CYP3A) was assumed to be 0.93, with the exception of alprazolam in which a value corrected for the percentage excreted unchanged in urine of 20% (Garzone and Kroboth, 1989) was accounted for yielding a value for f_m(CYP3A) of 0.74. The 0.93 values were estimated by assuming that 400 mg/day of ketoconazole causes complete inhibition of CYP3A in vivo and that the remaining capability to clear these drugs is via other processes. The effect of CYP3A inhibitors on the intestinal first-pass intrinsic clearance was estimated from the following,

(3)

in which the term [I]_g refers to the estimated concentration of inhibitor in the intestinal wall during the absorption phase and was estimated as the dose given at each occurrence (i.e., dose/day divided by the times dosed per day) using a value of 248 ml/min for Q_g (Rostami-Hodjegan and Tucker, 2004).

(4)

The term CL_int,g,inh/CL_int,g was entered into the following expression for intestinal extraction ratio (Wang et al., 2004).

(5)

The effect on hepatic intrinsic clearance was predicted as it was done for the other P450s using values for [I]_{in vivo} of systemic or estimated hepatic inlet, total or unbound (eq. 1), and the effects on intestinal and hepatic CYP3A-mediated metabolism were combined as a product (Wang et al., 2004).

(6)

Other Calculations. Mean -fold errors of prediction methods were calculated as follows:

(7)

and

(8)

as measures of bias and precision, respectively.

	Results

Top Abstract Materials and Methods Results Discussion Conclusions References

 
Overall Trends
In vitro inhibition data for all five P450 enzymes are listed in Tables 1 and 2. Drugs with the lowest IC₅₀ values included ketoconazole, itraconazole, ritonavir, quinidine, terbinafine, and fluvoxamine, which all possessed values below 0.1 µM for one of the P450 enzymes. The corresponding in vivo drug interaction data with references are listed in supplemental data.

A simple relationship between inhibitory potency and the magnitude of the DDI is presented in Fig. 1. With two exceptions (disulfiram and dicumarol), all inhibitors possessing in vitro potency values (IC₅₀) below 1 µM demonstrate drug interactions of at least 2-fold, a cutoff that has been designated as being potentially clinically relevant (Tucker et al., 2001). However, there were some inhibitors that did not demonstrate high in vitro potency yet still had been shown to cause drug interactions of greater than 2-fold. These include several CYP3A inhibitors known to cause irreversible inactivation (e.g., clarithromycin, erythromycin, and diltiazem) as well as some drugs that attain high concentrations in vivo (e.g., fluconazole and cimetidine). Thus, although in vitro potency can provide some indication of the potential to cause drug interactions in vivo, there are clearly exceptions.

View larger version (23K): [in this window] [in a new window]   Fig. 1. Bar graph of the magnitude of drug interactions versus in vitro inhibitory potency.

Quantitative Prediction of DDI from in Vitro Data: CYP1A2, CYP2C9, CYP2C19, and CYP2D6
Comparisons of in vivo DDI to predictions of DDI made from in vitro data for CYP1A2, CYP2C9, CYP2C19, and CYP2D6 are listed in Tables 4, 5, 6, 7, respectively.

CYP1A2 Interactions. Of the 22 drugs examined, fluvoxamine and enoxacin caused the largest increase in exposure to a CYP1A2-cleared drug (a 3.3- and 3.7-fold increase in theophylline; Wijnands et al., 1986; Rasmussen et al., 1997). The interaction with fluvoxamine was well predicted from in vitro data (Table 4), but that for enoxacin was not. Two-fold effects caused by propranolol and troleandomycin would not have been expected from in vitro data. Using systemic (free or total) or free hepatic inlet concentrations yielded the highest accuracy of identifying drugs that could cause a 2-fold increase in CYP1A2-cleared drugs (Table 4). It is interesting to note that CYP1A2 interactions with theophylline and clozapine would be grossly overestimated in some cases if the f_m(CYP1A2) values were not included in the predictions.

CYP2C9 Interactions. Twenty drugs had been tested for the effects on warfarin or tolbutamide pharmacokinetics, with the only drugs shown to cause a greater than 2-fold increase being fluconazole and sulfaphenazole (Table 5). Fluconazole interactions were predicted to be 5- to 6-fold, depending on whether systemic or estimated hepatic inlet concentrations were used; these predictions were greater than the 4.3-fold decrease in (S)-7-hydroxy-warfarin observed with coadministration with 300 mg of fluconazole. For sulfaphenazole, predicted interactions were mostly independent of which value for [I]_{in vivo} was used (4.5-fold) and were very close to the actual interaction (5.0-fold). This is consistent with a previous report of predictions of drug interactions from in vitro data for sulfaphenazole (Komatsu et al., 2000).

CYP2C19 Interactions. Of the drugs examined, 12 were examined in omeprazole or mephenytoin interaction studies. Of these, four had demonstrated interactions resulting in a 2-fold or greater magnitude. Three of these, fluconazole, fluvoxamine, and ticlopidine, would have been predicted from in vitro data, whereas the 2.1-fold increase in omeprazole exposure upon coadministration of clarithromycin would not have been expected because clarithromycin demonstrated little to no effect on CYP2C19 activity in vitro (Table 6). Although this could be due to an effect of clarithromycin on the CYP3A-catalyzed pathway of omeprazole metabolism, this is not likely because ketoconazole, which also potently inhibits CYP3A, did not cause a marked effect on omeprazole pharmacokinetics (Bottiger et al., 1997).

CYP2D6 Interactions. Fourteen drugs in this dataset had been tested as inhibitors of CYP2D6 in vivo (Table 7). Of these, the prediction of interactions greater than 2-fold was scattered. The 10-fold increase in desipramine caused by fluoxetine was not well predicted. However, in other work, we have shown that fluoxetine demonstrates substantial nonspecific binding in vitro, even at low concentrations of microsomes such as that used in this study, which can yield overestimates of inhibition constants (Margolis and Obach, 2003). Correction for this factor can substantially improve the prediction accuracy. For paroxetine, inhibition is known to be irreversible and mechanism-based (Bertelsen et al., 2003), and application of the equations for reversible inhibition would be inappropriate (see Discussion). In fact, in a recent analysis, application of a prediction model that accounts for mechanism-based inhibition yields an excellent prediction of drug interactions caused by paroxetine on CYP2D6-cleared drugs (Venkatakrishnan and Obach, 2005). Terbinafine and quinidine were the other known CYP2D6 inhibitors tested, and the in vitro data yielded good predictions of the magnitude of effect on desipramine exposure caused by these drugs.

Overall, these data suggest that, for P450s predominantly residing in the liver, in vitro data can be used to predict the magnitude of drug interactions in vivo. Exceptions to this include those drugs known to be mechanism-based inactivators, which would be expected to require appropriate in vitro inactivation data for accurate prediction rather than simple reversible inhibition data. The average -fold error of prediction of drug interaction magnitude ranged from 1.33- to 1.52-fold (Table 8), irrespective of which value was used for [I]_{in vivo}; however, this could be misleading because many of the interactants tested demonstrated less than 2-fold increases. If only those interactants causing a greater than 2-fold increase are included, the mean -fold error ranges from 1.70 to 2.28. These measures of bias (MFE) along with a comparison of precision values (RMSE) support the use of estimated free hepatic inlet C_max as the best value used for [I]_{in vivo} (Table 8); however, total systemic C_max performed almost equally well. The use of total hepatic inlet C_max or unbound systemic C_max generally yielded over- and underpredictions, respectively. Plots of predicted DDI using various values for [I]_{in vivo} versus actual DDI are in Fig. 2.

View larger version (29K): [in this window] [in a new window]   Fig. 2. Predicted magnitude of DDI versus actual DDI magnitude. The values used for [I]in vivo were total systemic Cmax (top left), unbound systemic Cmax (top right), estimated total hepatic inlet Cmax (bottom left), and estimated unbound hepatic inlet Cmax (bottom right). Symbols: , reversible inhibitors; fl, known mechanism-based inactivators; , inhibitors with known inhibitory metabolites. Dashed lines represent the boundaries of 2-fold interactions and the solid diagonal line is a line of unity.

Quantitative Prediction of DDI from in Vitro Data: CYP3A
Table 9 lists the in vitro inhibition data for 42 drugs for three CYP3A activities along with predictions of in vivo interactions and the actual interaction. The most potent of the three in vitro interactions was used to make the prediction. As with the other P450s, four possibilities for [I]_{in vivo} were tested for the component of hepatic inhibition and combined with estimated inhibitor concentrations in the intestine, yielding four predictions of CYP3A interactions. Of the 42 drugs, there were 16 that caused interactions of 2-fold or greater and 26 that did not. Several of these interactions were profound, such as 18x, 17x, and 20x interactions caused by itraconazole, ketoconazole, and ritonavir, respectively.

Four values for [I]_{in vivo} were used for prediction of the hepatic CYP3A component and combined with the intestinal component in predicting the magnitude of CYP3A-mediated interactions. Of these four, use of the unbound estimated hepatic inlet C_max values was generally most accurate (Table 8). Whereas mean -fold errors (bias) were similar for predictions made using total systemic C_max or unbound hepatic inlet C_max, the RMSE values (precision) for prediction showed an advantage of using the unbound hepatic inlet C_max for the prediction. These latter values were superior to those obtained when using other values for [I]_{in vivo}. In addition, ignoring the contribution of inhibition of intestinal CYP3A in the prediction of the magnitude of drug interactions yielded less accurate predictions; the mean -fold error for interactions more than two times rose from 1.87 to 2.61, and the RMSE increased from 4.13 to 5.76 when using unbound hepatic inlet C_max for [I]_{in vivo}. When the total hepatic inlet C_max value was used for [I]_{in vivo}, ignoring the contribution of the intestine was not detrimental, probably because of two errors in calculation compensating each other; i.e., employing an unrealistically high value for [I]_{in vivo} available to inhibit the hepatic CYP3A balanced by ignoring the effect on intestinal CYP3A.

The use of estimated total hepatic inlet concentrations clearly yielded many gross over-predictions of the magnitude of interaction, and this approach is not recommended. However, other approaches, such as the use of estimated free hepatic inlet concentrations, which performed reasonably with the other P450 enzymes, failed to predict several important CYP3A interactions. Many of the drugs that cause interactions with CYP3A substrates are mechanism-based inactivators, such as clarithromycin and diltiazem (Jones et al., 1999; Mayhew et al., 2000; Ito et al., 2003), erythromycin (Ito et al., 2003; McConn et al., 2004), mibefradil (Prueksaritanont et al., 1999), nefazodone (Kalgutkar et al., 2005), nelfinavir, ritonavir, and saquinavir (Koudriakova et al., 1998; Ernest et al., 2005), and verapamil (Wang et al., 2004). With some exceptions, drug interactions caused by these inactivators tended to be underpredicted using estimates of free hepatic inlet concentrations. In addition, for itraconazole, an unusual phenomenon, the inhibition of CYP3A caused by metabolites prior to release, has been reported to confound the prediction of the magnitude of drug interactions caused by this drug (Isoherannen et al., 2004). Drugs that do not cause CYP3A interactions in vivo were well predicted using systemic C_max or free hepatic inlet concentrations. The most successful approach overall for predicting CYP3A interactions was through the use of estimated free hepatic inlet C_max concentrations. However, this is limited because reversible inhibition data for known mechanism-based inactivators was used in the predictions.

A comparison of inhibition of testosterone, midazolam, and felodipine activities was also made. Many of the inhibitors showed a large difference in inhibitory potency for the different activities. Most notable was cyclosporin, for which IC₅₀ values of 4.5, 1.4, and 300 µM for testosterone, midazolam, and felodipine activities were measured, respectively (Table 2). Other drugs for which there was at least a 4-fold spread between the highest and lowest IC₅₀ values for CYP3A inhibition included atorvastatin, buspirone, erythromycin, nefazodone, omeprazole, quinidine, fluoxetine, clarithromycin, diltiazem, troleandomycin, itraconazole, and roxithromycin. Of these drugs that showed such a difference, midazolam hydroxylase seemed to be most frequently the most potently inhibited activity followed by testosterone and felodipine. Whereas there were a few cases in which felodipine was the most potently inhibited activity (e.g., saquinavir), in none of these cases was the IC₅₀ for felodipine activity uniquely low compared with the other two. Although some differences can be observed in in vitro inhibition potency for the three CYP3A activities, there were very few cases in which the in vitro CYP3A inhibition data yielded predictions of drug interaction magnitudes that would be interpreted differently. Clearly, if felodipine activity were the one used to judge whether cyclosporine would cause drug interactions with CYP3A-cleared drugs, it would be predicted that no interaction would occur (AUC_i/AUC = 1.02), although a 2.55-fold interaction with simvastatin caused by cyclosporine has been reported. However, the use of the IC₅₀ for midazolam using the unbound hepatic inlet or total systemic concentration of cyclosporine yields a predicted increase of 1.7- to 2.5-fold for simvastatin exposure (Table 9). The only other drugs for which IC₅₀ values were different for the three CYP3A activities, such that it would be possible to predict a different outcome for in vivo drug interactions (i.e., interaction of >2x versus <2x), were fluvoxamine, erythromycin, nefazodone, and mibefradil. Overall, midazolam 1'-hydroxylase and testosterone 6-hydroxylase seemed to be slightly more predictive of CYP3A-mediated drug interactions than felodipine dehydrogenase (Table 10).

	Discussion

Top Abstract Materials and Methods Results Discussion Conclusions References

 
In the past 15 years, our understanding of the human cytochrome P450 enzymes has increased to the point where the mechanistic underpinnings of drug-drug interactions are well understood. Specific drug metabolism reactions are attributed to specific P450 enzymes, and several drugs have been identified as having a substantial fraction of total clearance catalyzed by a single enzyme. Gathering in vitro inhibition data using human liver microsomes has become a commonplace activity in the search for new drugs that will not cause serious pharmacokinetic drug-drug interactions. However, to date, a broadly applicable method whereby quantitative predictions of the magnitude of drug interactions can be made has remained elusive. A general observation that drugs that have K_i values below 1 µM tend to cause DDI in vivo has been made (Wrighton et al., 2000). Some reports exist wherein a correlation was made between in vivo DDI and in vitro inhibition data for one or a few drugs. However, broad and comprehensive analyses have not been routinely reported. Ito et al. (2004) compared a literature database of in vitro inhibition data and in vivo DDI data and found a general correlation, but actual predictions of the magnitudes of DDI were not made.

The greatest uncertainty in predicting the magnitude of in vivo drug interactions resides in the values used for [I]_{in vivo}. The most appropriate value would be the concentration available to the enzyme in the liver; however such a value cannot be determined in vivo. Possibilities of surrogate in vivo concentrations that could be used include the free or total systemic concentrations, or free or total hepatic inlet concentrations estimated to occur during the absorption phase after oral administration. Classic pharmacologic theory would favor the "unbound drug hypothesis"; drawing analogy to P450 enzymes as pharmacological targets no different from others. However, others have pointed toward the usefulness of estimated total hepatic concentrations as being of use in making in vitro-in vivo correlations for metabolic drug interactions (Von Moltke et al., 1994). The possibility of discordance between free systemic (or hepatic inlet) and free intrahepatic concentrations arising from active uptake processes, which would complicate the prediction of magnitudes of drug interactions from in vitro inhibition data, is likely to be important; however, incorporating such a possibility was beyond the scope of this treatment. In the present analysis, systemic and estimated hepatic inlet concentrations, along with free and total concentrations, were considered for [I]_{in vivo} (Fig. 2). Overall, the use of estimated unbound hepatic inlet C_max during the absorptive phase yielded the most accurate predictions of the magnitudes of DDI. However, even this likely represents an oversimplification of the actual situation because concentrations of both substrate and inhibitor would be constantly changing over time, and for many inhibitors, active uptake and efflux processes in hepatocytes will influence the intracellular concentration of the inhibitor.

Unlike CYP1A2, CYP2C9, CYP2C19, and CYP2D6, CYP3A offers other complicating factors. First, CYP3A activities measured in human liver microsomes actually comprise a composite of two enzymes: CYP3A4 and CYP3A5. Whereas some work has been reported regarding the activities of these two enzymes for various CYP3A substrates, a clear understanding of the contribution of each in pooled human liver microsomes is not yet available. Second, from a comprehensive dataset, it has been proposed that CYP3A4 possesses three substrate binding types (Kenworthy et al., 1999), such that inhibitors may act differently toward some CYP3A substrates and not others. From this phenomena, the question arises whether in vitro inhibition data gathered using one substrate are different from data gathered using another with regard to prediction of in vivo drug interactions. To address this question, inhibition data were gathered for the same set of inhibitors on three CYP3A activities: testosterone 6-hydroxylase, midazolam 1'-hydroxylase, and felodipine dehydrogenase activities, which represent the aforementioned three substrate binding types. Third, CYP3A4 is present in the intestine, and this tissue has been demonstrated to contribute a substantial portion to first-pass extraction of some CYP3A-cleared drugs. The methods employed for predicting drug interactions from in vitro data for the other P450 enzymes only considered the liver. However, the in vivo probe substrates (midazolam, triazolam, alprazolam, nifedipine, simvastatin, and buspirone) all have different relative contributions of gut and liver to exposure after oral administration. By applying eq. 6 (Wang et al., 2004), the effect of inhibitors on CYP3A in the intestine can be included for drugs like midazolam and buspirone, which undergo considerable intestinal metabolism after oral administration, whereas the effect on intestinal metabolism for drugs like alprazolam will be blunted because the intestine plays a more minor role than the liver in the metabolic clearance of alprazolam. This is important when considering CYP3A-cleared drugs.

Accounting for these complexities, particularly the effect of the intestine, seemed to improve the DDI predictions for CYP3A inhibitors. However, several were still underpredicted, probably because of the fact that several are CYP3A mechanism-based inactivators. With some exceptions, drug interactions caused by these inactivators tended to be underpredicted using estimates of free hepatic inlet concentrations. For mechanism-based inactivators, it has been shown that other in vitro drug interaction parameters (i.e., k_inact and K_I) and equations are necessary (Mayhew et al., 2000; Wang et al., 2004). Preliminary data have shown that a simple measurement of IC₅₀ after a 30-min preincubation of inactivator with microsomes and NADPH yields values that predict the magnitude of drug interactions caused by mechanism-based inactivators (data not shown). A comprehensive investigation of inactivators (for CYP3A and other P450 enzymes) and prediction of drug interactions is presently underway and will be reported in due course.

A comprehensive in vitro assessment of substrate specificity of CYP3A inhibitor effects (Kenworthy et al., 1999) has provided evidence in support of the concept of different testosterone-like, dihydropyridine-like, and benzodiazepine-like substrate classes. Consistently, meaningful differences in potency among the three index reactions for several inhibitors were observed in our study. However, in this reasonably sized dataset (n = 42 drugs with 16 drugs producing a >2-fold interaction magnitude), it was not clear that one substrate provided substantially superior performance in predicting DDI (Table 10). Analysis of the DDI predictions made from in vitro data from the three different substrates suggested different conclusions depending on the statistical approach employed. For example, if categorizing DDI based on a 2-fold cutoff, midazolam data provided the greatest percentage of success followed by testosterone and then felodipine. However, if using MFE, testosterone was best followed by midazolam and then felodipine, and if using RMSE, testosterone was best but midazolam was worst. Because these data are limited to 16 drugs for which >2x interactions are observed and because many of these are mechanism-based inactivators, it cannot be concluded that one CYP3A activity is vastly superior to the others; therefore, continued monitoring of the prediction accuracy of the three in vitro activities is warranted. This is similar to other recently reported findings (Galetin et al., 2005).

	Conclusions

Top Abstract Materials and Methods Results Discussion Conclusions References

 
The following overall conclusions can be made from this analysis. 1) A crude comparison of in vitro inhibition potency and magnitude of drug interactions demonstrated that, in almost all cases, if the inhibitory potency was 1 µM, an in vivo drug-drug interaction would be observed; however, if the inhibitory potency was >10 µM, there still is the possibility that the drug could cause an interaction (Fig. 1). Thus, predicting drug-drug interactions from in vitro potency data alone is not advocated. 2) For any given drug, the order of potency values for different P450 enzymes generally lines up with the magnitude of DDI in vivo with selective probe substrates for these enzymes. Of 21 drugs, 18 demonstrated this relationship. 3) Through the use of eq. 1, which includes f_m(CYP) as an important component (Ito et al., 2005), in vitro inhibition data can be reliably used to identify those drugs likely to cause in vivo drug interactions with CYP1A2, CYP2C9, CYP2C19, and CYP2D6. However, this requires several simplifications: an assumption that the estimated unbound hepatic inlet C_max value is the most meaningful in vivo concentration to use in the equation; the mechanism of inhibition is not irreversible mechanism-based inactivation; there are no metabolites of the inhibitors that also possess inhibitory activity; and that the magnitude of nonspecific binding in vitro under the conditions used to measure the marker substrate activities is negligible and hence does not affect the measurement of inhibitory potency value. Furthermore, our analysis has ignored the potential impact of active uptake and efflux by the liver because the in vitro system used was liver microsomes. For some compounds, the interplay between metabolism and transport would be expected to have an impact on the magnitude of DDI if the drugs under consideration are substrates/inhibitors for both processes. It could be proposed that intact hepatocytes would be a superior system to use for prediction of DDI from in vitro inhibition experiments, because this system would have uptake/efflux mechanisms. However the data presented in this report suggest that liver microsomal systems yield acceptable inhibition data for prediction of DDI as shown in the plot of the predicted DDI (using unbound hepatic inlet C_max) versus actual DDI in Fig. 2. 4) For CYP3A inhibitors, the effects on both hepatic and intestinal metabolism should be considered. Although differences in in vitro inhibitory potency were observed for different CYP3A substrates for many of the inhibitors tested in this study, few of these resulted in a different prediction of the magnitude of CYP3A-mediated drug interactions.

	Acknowledgements

 
We thank Dr. Aleksandra Galetin (University of Manchester) for useful discussions of the CYP3A inhibition data.

	Footnotes

 
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.

doi:10.1124/jpet.105.093229.

ABBREVIATIONS: P450, cytochrome P450; DDI, drug-drug interaction(s); MFE, mean-fold error; RMSE, root mean-squared error; AUC, area under the curve.

The online version of this article (available at http://jpet.aspetjournals.org) contains supplemental material.

Address correspondence to: Dr. R. Scott Obach, Pfizer Global Research and Development, Groton Laboratories, MS4088, Groton, CT 06340. E-mail: r.scott.obach{at}pfizer.com

	References

Top Abstract Materials and Methods Results Discussion Conclusions References

 

Benet LZ, Cummins CL, and Wu CY (2003) Transporter-enzyme interactions: implications for predicting drug-drug interactions from in vitro data. Curr Drug Metab 4: 393-398.[CrossRef][Medline]

Bertelsen KM, Venkatakrishnan K, Von Moltke LL, Obach RS, and Greenblatt DJ (2003) Apparent mechanism-based inhibition of human CYP2D6 in vitro by paroxetine: comparison with fluoxetine and quinidine. Drug Metab Dispos 31: 289-293.[Abstract/Free Full Text]

Bjornsson TD, Callaghan JT, Einolf HJ, Fischer V, Gan L, Grimm S, Kao J, King SP, Miwa G, Ni L, et al. (2003) The conduct of in vitro and in vivo drug-drug interaction studies: a Pharmaceutical Research and Manufacturers of America (PhRMA) perspective. Drug Metab Dispos 31: 815-832.[Abstract/Free Full Text]

Blanchard N, Richert L, Coassolo P, and Lave T (2004) Qualitative and quantitative assessment of drug-drug interaction potential in man, based on K_i, IC₅₀ and inhibitor concentration. Curr Drug Metab 5: 147-156.[CrossRef][Medline]

Bottiger Y, Tybring G, Gotharson E, and Bertilsson L (1997) Inhibition of the sulfoxidation of omeprazole by ketoconazole in poor and extensive metabolizers of S-mephenytoin. Clin Pharmacol Ther 62: 384-391.[CrossRef][Medline]

Brosen K, Hansen JG, Nielsen KK, Sindrup SH, and Gram LF (1993) Inhibition by paroxetine of desipramine metabolism in extensive but not in poor metabolizers of sparteine. Eur J Clin Pharmacol 44: 349-355.[CrossRef][Medline]

Chien JY, Mohutsky MA, and Wrighton SA (2003) Physiological approaches to the prediction of drug-drug interactions in study populations. Curr Drug Metab 4: 347-356.[CrossRef][Medline]

Dain JG, Nicoletti J, and Ballard F (1997) Biotransformation of clozapine in humans. Drug Metab Dispos 25: 603-609.[Abstract/Free Full Text]

Ernest CS, Hall SD, and Jones DR (2005) Mechanism-based inactivation of CYP3A by HIV protease inhibitors. J Pharmacol Exp Ther 312: 583-591.[Abstract/Free Full Text]

Galetin A, Ito K, Hallifax D, and Houston JB (2005) CYP3A4 substrate selection and substitution in the prediction of potential drug-drug interactive. J Pharmacol Exp Ther 314: 180-190.[Abstract/Free Full Text]

Garzone PD and Kroboth PD (1989) Pharmacokinetics of the newer benzodiazepines. Clin Pharmacokinet 16: 337-364.[Medline]

Houston JB and Galetin A (2003) Progress towards prediction of human pharmacokinetic parameters from in vitro technologies. Drug Metab Rev 35: 393-415.[CrossRef][Medline]

Huang SM, Lesko LJ, and Williams RL (1999) Assessment of the quality and quantity of drug-drug interaction studies in recent NDA submissions: study design and data analysis issues. J Clin Pharmacol 39: 1006-1014.[Abstract]

Isoherannen N, Kunze KL, Allen KE, Nelson WL, and Thummel KE (2004) Role of itraconazole metabolites in CYP3A4 inhibition. Drug Metab Dispos 32: 1083-1091.[Abstract/Free Full Text]

Ito K, Brown HS, and Houston JB (2004) Database analyses for the prediction of in vivo drug-drug interactions from in vitro data. Br J Clin Pharmacol 57: 473-486.[CrossRef][Medline]

Ito K, Hallifax D Obach RS, and Houston JB (2005) Impact of parallel pathways of drug elimination and multiple CYP involvement on drug-drug interactions: CYP2D6 paradigm. Drug Metab Dispos 33: 837-844.[Abstract/Free Full Text]

Ito K, Ogihara K, Kanamitsu S, and Itoh T (2003) Prediction of the in vivo interaction between midazolam and macrolides based on in vitro studies using human liver microsomes. Drug Metab Dispos 31: 945-954.[Abstract/Free Full Text]

Jones DR, Gorski JC, Hamman MA, Mayhew BS, Rider S, and Hall SD (1999) Diltiazem inhibition of cytochrome P-450 3A activity is due to metabolite intermediate complex formation. J Pharmacol Exp Ther 290: 1116-1125.[Abstract/Free Full Text]

Kalgutkar AS, Vaz ADN, Lame ME, Henne KR, Soglia S, Zhao SX, Abramov YA, Lombardo F, Collin C, Hendsch ZS, et al. (2005) Bioactivation of the nontricyclic antidepressant nefazodone to a reactive quinine-imine species in human liver microsomes and recombinant cytochrome P4503A4. Drug Metab Dispos 33: 243-253.[Abstract/Free Full Text]

Kanamitsu SI, Ito K, and Sugiyama Y (2000) Quantitative prediction of in vivo drug-drug interactions from in vitro data based on physiological pharmacokinetics: use of maximum unbound concentration of inhibitor at the inlet to the liver. Pharm Res (NY) 17: 336-343.

Kenworthy KE, Bloomer JC, Clarke SE, and Houston JB (1999) CYP3A4 drug interactions: correlation of 10 in vitro probe substrates. Br J Clin Pharmacol 48: 716-727.[CrossRef][Medline]

Kirchheiner J, Bauer S, Meineke I, Rohde W, Prang V, Meisel C, Roots I, and Brockmöller J (2002) Impact of CYP2C9 and CYP2C19 polymorphisms on tolbutamide kinetics and the insulin and glucose response in healthy volunteers. Pharmacogenetics 12: 101-109.[CrossRef][Medline]

Komatsu K, Ito K, Nakajima Y, Kanamitsu SI, Imaoka S, Funae Y, Green CE, Tyson CA, Shimada N, and Sugiyama Y (2000) Prediction of in vivo drug-drug interactions between tolbutamide and various sulfonamides in humans based on in vitro experiments. Drug Metab Dispos 28: 475-481.[Abstract/Free Full Text]

Koudriakova T, Iatsimirskaia E, Utkin I, Gangl E, Vouros P, Storozhuk E, Orza D, Marinina J, and Gerber N (1998) Metabolism of the human immunodeficiency virus protease inhibitors indinavir and ritonavir by human intestinal microsomes and expressed cytochrome P4503A4/3A5: mechanism-based inactivation of cytochrome P4503A by ritonavir. Drug Metab Dispos 26: 552-561[Abstract/Free Full Text]

Kupfer A, Desmond P, Patwardhan R, Schenker S, and Branch RA (1984) Mephenytoin hydroxylation deficiency: kinetics after repeated doses. Clin Pharmacol Ther 35: 33-39.[Medline]

Lennard MS, Tucker GT, Silas JH, Freestone S, Ramsay LE, and Woods HF (1983) Differential stereoselective metabolism of metoprolol in extensive and poor debrisoquin metabolizers. Clin Pharmacol Ther 34: 732-737.[Medline]

Margolis JM and Obach RS (2003) Impact of nonspecific binding to microsomes and phospholipid on the inhibition of cytochrome P4502D6: Implications for relating in vitro inhibition data to in vivo drug interactions. Drug Metab Dispos 31: 606-611.[Abstract/Free Full Text]

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: 1031-1037.[Abstract/Free Full Text]

McConn DJ, Lin YS, Allen K, Kunze KL, and Thummel KE (2004) Differences in the inhibition of cytochromes P450 3A4 and 3A5 by metabolite-inhibitor complex-forming drugs. Drug Metab Dispos 32: 1083-1091.[Abstract/Free Full Text]

Monks TJ, Caldwell J, and Smith RL (1979) Influence of methylxanthine-containing foods on theophylline metabolism and kinetics. Clin Pharmacol Ther 26: 513-524.[Medline]

Neal JM, Kunze KL, Levy RH, O'Reilly RA, and Trager WF (2003) Kiiv, an in vivo parameter for predicting the magnitude of a drug interaction arising from competitive enzyme inhibition. Drug Metab Dispos 31: 1043-1048.[Abstract/Free Full Text]

Prueksaritanont T, Ma B, Tang C, Meng Y, Assang C, Lu P, Reider PJ, Lin JH, and Baillie TA (1999) Metabolic interactions between mibefradil and HMG-CoA reductase inhibitors: an in vitro investigation with human liver preparations. Br J Clin Pharmacol 47: 291-298.[CrossRef][Medline]

Rasmussen BB, Jeppesen U, Gaist D, and Brosen K (1997) Griseofulvin and fluvoxamine interactions with the metabolism of theophylline. Ther Drug Monit 19: 56-62.[CrossRef][Medline]

Rostami-Hodjegan A and Tucker GT (2004) "In silico" simulations to assess the "in vivo" consequences of "in vitro" metabolic drug-drug interactions. Drug Discov Today Tech 1: 441-448.

Rowland M and Matin SB (1973) Kinetics of drug-drug interactions. J Pharmacokinet Biopharm 1: 553-567.[CrossRef]

Scordo MG, Pengo V, Spina E, Dahl ML, Gusella M, and Padrini R (2002) Influence of CYP2C9 and CYP2C19 genetic polymorphisms on warfarin maintenance dose and metabolic clearance. Clin Pharmacol Ther 72: 702-710.[CrossRef][Medline]

Shou M (2005) Prediction of pharmacokinetics and drug-drug interactions from in vitro metabolism data. Curr Opin Drug Disc Dev 8: 66-77.[Medline]

Thummel KE, Kunze KL, and Shen DD (1997) Enzyme-catalyzed processes of first-pass hepatic and intestinal drug extraction. Adv Drug Del Rev 27: 99-127.[CrossRef][Medline]

Tucker GT, Houston JB, and Huang S-M (2001) Optimising drug development: Strategies to assess drug metabolism/transporter interaction potential—toward a consensus. Clin Pharmacol Ther 70: 103-114.[CrossRef][Medline]

Venkatakrishnan K and Obach RS (2005) In vitro-in vivo extrapolation of CYP2D6 inactivation by paroxetine: prediction of nonstationary pharmacokinetics and drug interaction magnitude. Drug Metab Dispos 33: 845-852.[Abstract/Free Full Text]

Venkatakrishnan K, von Moltke LL, Obach RS, and Greenblatt DJ (2003) Drug metabolism and drug interactions: application and clinical value of in vitro models. Curr Drug Metab 4: 423-459.[CrossRef][Medline]

Von Moltke LL, Greenblatt DJ, Cotreau-Bibbo MM, Duan SX, Harmatz JS, and Shader RI (1994) Inhibition of desipramine hydroxylation in vitro by serotonin-reuptake-inhibitor antidepressants and by quinidine and ketoconazole: a model system to predict drug interactions in vivo. J Pharmacol Exp Ther 268: 1278-1283.[Abstract/Free Full Text]

Walsky RL and Obach RS (2004) Validated assays for human cytochrome P450 activities. Drug Metab Dispos 32: 647-660.[Abstract/Free Full Text]

Wang YH, Jones DR, and Hall SD (2004) Prediction of cytochrome P450 3A inhibition by verapamil enantiomers and their metabolites. Drug Metab Dispos 32: 259-266.[Abstract/Free Full Text]

Wijnands WJ, Vree TB, and van Herwaarden CL (1986) Influence of quinolone derivatives on theophylline clearance. Br J Clin Pharmacol 22: 677-683.[Medline]

Wrighton SA, Schuetz EG, Thummel KE, Shen DD, Korzekwa KR, and Watkins PB (2000) The human CYP3A subfamily: practical considerations. Drug Metab Rev 32: 339-361.[CrossRef][Medline]

Wu CY and Benet LZ (2005) Predicting drug disposition via application of BCS: transport/absorption/elimination interplay and development of a biopharmaceutics drug disposition classification system. Pharm Res (NY) 22: 11-23.

Yang J, Tucker GT, and Rostami-Hodjegan A (2004) Cytochrome P450 3A expression and activity in the human small intestine. Clin Pharmacol Ther 76: 391.[CrossRef][Medline]

Yao C and Levy RH (2002) Inhibition-based metabolic drug-drug interactions: predictions from in vitro data. J Pharm Sci 91: 1923-1935.[CrossRef][Medline]

Yasui-Furukori N, Takahata T, Nakagami T, Yoshiya G, Inoue Y, Kaneko S, and Tateishi T (2004) Different inhibitory effect of fluvoxamine on omeprazole metabolism between CYP2C19 genotypes. Br J Clin Pharmacol 57: 487-494.[CrossRef][Medline]

This article has been cited by other articles:

				Y. Kapelyukh, M. J. I. Paine, J.-D. Marechal, M. J. Sutcliffe, C. R. Wolf, and G. C. K. Roberts Multiple Substrate Binding by Cytochrome P450 3A4: Estimation of the Number of Bound Substrate Molecules Drug Metab. Dispos., October 1, 2008; 36(10): 2136 - 2144. [Abstract] [Full Text] [PDF]

				D. Aidasani, M. J. Zaya, P. B. Malpas, and C. W. Locuson In Vitro Drug-Drug Interaction Screens for Canine Veterinary Medicines: Evaluation of Cytochrome P450 Reversible Inhibition Drug Metab. Dispos., August 1, 2008; 36(8): 1512 - 1518. [Abstract] [Full Text] [PDF]

				K. Grime, P. J. H. Webborn, and R. J. Riley Functional Consequences of Active Hepatic Uptake on Cytochrome P450 Inhibition in Rat and Human Hepatocytes Drug Metab. Dispos., August 1, 2008; 36(8): 1670 - 1678. [Abstract] [Full Text] [PDF]