Mini-series: I. Basic science. Uncertainty and inaccuracy of predicting CYP-mediated in vivo drug interactions in the ICU from in vitro models: focus on CYP3A4
- First Online:
- Cite this article as:
- Mouly, S., Meune, C. & Bergmann, J. Intensive Care Med (2009) 35: 417. doi:10.1007/s00134-008-1384-1
- 303 Views
Drug–drug interactions (DDIs) contribute significantly to the incidence of adverse drug reactions. Important advances in the knowledge of human drug-metabolizing enzymes have fueled the integration of in vitro drug metabolism and clinical DDIs studies for use in drug development programs and in the clinical setting. The activity of cytochrome P450 (CYP) 3A4 and P-glycoprotein are critical determinant of drug clearance, interindividual variability in drug disposition and clinical efficacy, and appears to be involved in the mechanism of numerous clinically relevant DDIs. Cell-based in vitro models are being increasingly applied in elucidating the pharmacokinetic profile of drug candidates during the preclinical steps of drug development. Human liver, intestinal samples and recombinant human CYP3A4 are now readily available as in vitro screening tools to predict the potential for in vivo DDIs. Although it is easy to determine in vitro metabolic DDIs, the interpretation and extrapolation of in vitro interaction data to in vivo situations requires a good understanding of pharmacokinetic principles. Clinicians and pharmacokineticists should recognize that in vitro models may not be clinically relevant in all situations. In the current article, research will be presented on drug metabolism and DDIs along with examples illustrating the utility of specific in vitro or in vivo approaches. In addition, the impact and clinical relevance of complexities such as dosing-route dependent effects, multi-site kinetics of drug-metabolizing enzymes and non-CYP determinants of metabolic clearance will be addressed.
KeywordsDrug–drug interactionsCytochrome P450 3A4P-glycoproteinHuman liver microsomesCaco-2 cells
Intensive care unit
Human immunodeficiency virus
Human liver microsomes
Multidrug resistance related proteins
Human Pregnane X receptor
Human constitutive androstane receptor
Quantitative-drug interaction prediction system
Adverse drug reactions, defined as any noxious, unintended and undesired effect of a drug occurring at doses used for human prophylaxis, diagnosis or therapy, are a major source of complications in the intensive care unit (ICU) [1–4]. Concomitant administration of multiple drugs frequently causes severe pharmacokinetic or pharmacodynamic drug–drug interactions (DDIs) resulting in the possibility of enhanced toxicity and/or therapy failure. A DDI occurs when the effects of one drug are changed by the presence of another drug, food, or some environmental factor . For example, intravenous sildenanfil augmented the pulmonary vasodilator effects of inhaled nitric oxide in infants early after cardiac surgery but produced systemic hypotension and impaired oxygenation, which was not improved or even prevented by nitric oxide . Epinephrine, norepinephrine and dopamine infusions have been shown to decrease propofol concentrations by up to 63% during continuous infusion, thus reversing propofol anesthesia . Although the results of a given DDI can be either positive (e.g., using grapefruit juice to increase cyclosporine bioavailability) or negative (e.g., erythromycin plus cisapride increases the risk of torsades de pointes), there are often unpredicted and undesired. Usually, the risk of significant DDIs increases when a patient requires a large number of medications, larger doses of medication and a longer duration of therapy . There are three major categories of DDIs: pharmacodynamic, as shown above, pharmaceutical, and pharmacokinetic [6, 8, 9]. When there is physical incompatibility between two different drugs, a pharmaceutical interaction may occur . An example of this would be the administration of ciprofloxacin with calcium carbonate . The calcium in the antacid binds the ciprofloxacin, forming a complex that cannot be absorbed. In general, DDIs that affect a drug’s absorption, distribution (or protein binding), metabolism, or elimination are considered pharmacokinetic interactions. By understanding the mechanism by which drugs are normally metabolized, many pharmacokinetic DDIs can be better understood, increasing the ability to predict and prevent potentially serious interactions.
The daily practicing intensivist must remain aware of the major mechanisms for DDIs, among which the drug-metabolizing enzyme inhibitory or induction potential of associated chemical entities are paramount. Metabolism-based DDIs are indeed largely due to changes in levels or activity of drug-metabolizing enzymes caused by one drug, leading to changes in the systemic exposure clearance of another . Important advances in the knowledge of human drug-metabolizing enzymes have largely fueled the integration of in vitro drug metabolism and clinical DDIs studies for use in drug development programs and to a lesser extent in the clinical setting . The superfamily of cytochromes P450 (CYPs) plays a key role in this phenomenon, including in critically ill patients but, because of the complexity of interactions between drugs and isozymes, it becomes more and more difficult for clinicians to master the knowledge required to predict the occurrence of such DDIs . This variability may reflect enzyme inhibition or induction, along with now increasingly identified genetic differences in the expression of drug metabolizing enzymes. Many pharmacological and disease-related factors, e.g., extended use of proton pump inhibitors for gastric acid suppression, drug-related inhibition of intestinal motility, catecholamine-related circulatory changes impairing organ perfusions and altering drug volume of distribution, enteral versus parenteral feeding and their effect on hepatic and intestinal blood flow, morbid obesity, human immunodeficiency virus (HIV) infection, frequent renal and hepatic impairment, along with concomitant multiple drug regimens may account for the wide interindividual variability in drug exposure and response in critically ill patients and for the high risk for DDIs to occur [7, 13–15]. Early assessment of drug metabolism characteristics had been made possible by the advances made in analytical capabilities and in in vitro technologies that are employed to predict in vivo metabolite profile and DDI potential . The current basic science part of this “mini-series” will review these in vitro systems and the pharmacological basis for drug metabolism. It is the purpose of the authors to make this manuscript as practical as possible for daily practising intensivists, for a better understanding of the wide interindividual variability in drug exposure and response and for the high risk for DDIs in critically ill patients.
Clinical rationale for predicting adverse drug reactions and drug–drug interactions in intensive care medicine
ICU patients usually receive more therapies than patients on general medical and surgical wards and practitioners caring for these patients are presented with the challenge of monitoring each of these therapies for efficacy and toxicity . Thus an assessment of potential DDIs is imperative. In a study among 205 outpatients seen in an emergency room in Los Angeles, the prevalence of potential DDIs was 13% in those who received two medications, and 82% in those who received seven or more drugs . Age and its related physiologic changes, malnutrition, malabsorption, chronic liver disease, impaired renal function, the number of medications, long-term drug dosing, and pharmacogenetic characteristics of individual patients are risk factors for DDIs [5, 8]. Considering these factors, it is not surprising that critically ill patients are often at much greater risk of serious DDIs than are most other patient populations.
Because human liver and intestinal samples, cell-based in vitro models and recombinant human CYPs are now readily available, in vitro systems are being used as screening tools to predict the potential for in vivo DDIs . Although it is easy to determine in vitro metabolic DDIs, the interpretation and extrapolation of in vitro interaction data to in vivo situations require a good understanding of pharmacokinetic principles. Both clinicians and pharmacokineticists should keep in mind that in vitro models do not incorporate the disease-related factors cited above and are not clinically relevant in all situations. Ideally, prescribing should be based on the patient’s individual biotransformation capacity and the clinician should be ideally guided by bedside tools predicting drug interactions and altered CYP-activity.
Among the numerous drug-metabolizing enzymes expressed in humans, CYP3A4 is the major metabolic pathway taken by drugs in the human body. About 60% of currently marketed drugs and of new molecular entities are metabolized by CYP3A4 . CYP3A4 is both inhibited and induced by drugs, and this accounts for the majority of clinically relevant DDIs. The importance of CYP3A4 has been highlighted by the recent removal of terfenadine, cisapride and cerivastatin from the market due to life-threatening toxicity produced in patients receiving inhibitors of CYP3A4. These drugs are largely metabolized by CYP3A4; simultaneous administration with CYP3A4 inhibitors resulted in dangerous increases in drug blood levels. Conversely, mibefradil was withdrawn from the market because it was a potent inhibitor of CYP3A4 and could cause toxic reactions to some drugs largely metabolized by CYP3A4 (such as statins and anesthetic drugs).
In the current basic science article, in vitro approaches to drug metabolism and DDIs will be presented along with examples illustrating the utility of specific in vitro models. In addition, the impact and clinical relevance of complexities such as dosing-route dependent effects, multi-site kinetics of CYP3A4 and non-CYP determinants of metabolic clearance will be reviewed. Although the in vitro models and methods of prediction are numerous, there remain a number of unresolved factors that affect the accuracy of the predictions. These factors are also discussed to provide a caution for researchers performing prediction studies and for clinicians, as well, always looking for a predictive, reproducible, non invasive tool to optimize drug dosing and drug combinations on an individual basis. The purpose of the miniseries is to give an overview of the various drug metabolic pathways in humans and of the in vitro systems available to study clinically relevant DDIs. However, due to the predominant role of the largely expressed CYP3A4 in the metabolism of drugs frequently and concomitantly administered in ICU patients and in clinically relevant DDIs involving these drugs, the manuscript will focus on this pathway for clarity purpose and according to literature search and analysis.
Important advances in the knowledge of human drug-metabolizing enzymes
Normal drug metabolism
The process of drug metabolism normally causes medications to become less active but occasionally results in a more active and/or toxic product which may stay in the body longer than the parent compound . Common examples of this include the antiplatelet drug clopidrogel, the antihypertensive losartan and the analgesics codeine and acetaminophen [25–27]. In these situations, induction of drug metabolism will not accelerate drug elimination, but rather increase the risk for drug accumulation and toxicity.
Phase I enzymes: the cytochromes P450 (CYPs)
Phase I enzymes are primarily responsible for changing the existing drug structure in a relatively small but very important way. An example would be oxidation where a single oxygen molecule is incorporated into the drug. Phase I enzymes are generally present in smaller quantities than are most phase II enzymes. The location of phase I enzymes makes them extremely sensitive to even small changes that occur within the body, which explains why they are common targets for DDIs [5, 19]. As discussed earlier, most phase I metabolism is primarily controlled by the CYP enzymes. Cytochrome P450s constitute a superfamily of hemoproteins that are the terminal oxidases of the mixed function oxidase system [21, 28–30]. The CYP superfamily is comprised of 270 genes (with 18 recorded in mammals) found in 23 eukaryotes (both plants and animals) and 6 prokaryotes. The human genome contains 57 CYP genes and 33 pseudogenes arranged into 18 families and 42 subfamilies . In the recommended nomenclature, the deduced amino-acid sequences from the genes are compared and first divided into families, which are comprised of those CYPs that share at least 40% identity (currently 30 families are documented) [30, 31]. The mammalian families are divided further into subfamilies, which are comprised of those forms that are at least 55% related by their deduced amino-acid sequences. Only 3 CYP gene families (i.e., CYP1, CYP2 and CYP3) of the 30 families identified are thought to be responsible for the majority of drug metabolism.
Local anesthetic agent: ropivacaïne
Anti-Alzheimer agent: tacrine
Antiasthmatic agents: theophylline, zileuton
Andepressant agents: amitriptyline, clomipramine, fluvoxamine, imipramine
Antipsychotic agents: clozapine, haloperidol
Angiotensin II receptor blockers: irbesartan, losartan
Anticoagulant agent: warfarin
Anticonvulsant agent: phenytoin
Cancer chemotherapeutic agent: tamoxifen
Hypoglycemic oral agents: glipizide, glyburide, tolbutamide
Nonsteroidal anti-inflammatory drug (NSAIDs): celecoxib, diclofenac, ibuprofen, naproxen
Antidepressant agents: amitriptyline, citalopram, clomipramine, imipramine
Immunosuppressant agent: cyclophosphamide (prodrug)
Proton pump inhibitors: lansoprazole, omeprazole, pantopropazole
Analgesic agents, narcotic: codeine (prodrug), tramadol (prodrug)
Anesthetic agent, local: lidocaïne
Antidepressant agents: amitriptyline, clomapramine, desipramine, imipramine, paroxetine
Antipsychotic agents: haloperidol, perphenazine, risperidone
Beta blockers: carvedilol, metoprolol, propanolol, timolol
Analgesic agent, nonnarcotic: acetaminophen
Anesthetic agents, general: enflurane, halothane, isoflurane, sevoflurane
Muscle relaxant: chlorzoxazone
Recreational drug: ethanol
Analgesic agent, narcotic: alfentanil
Anesthetic agent, local: lidocaïne
Antibiotic agents: clarithromycin, erythromycin
Anticoagulant agent: warfarin
Anticonvulsant agent: carbamazepine
Antihistamine agents: astemizole, terfenadine
Antipsychotic agents: haloperidol, pimozide
Benzodiazepine agents: alprazolam, diazepam, midazolam, triazolam
Calcium channel-blocking agents: amlodipine, diltiazem, felodipine, nifedipine, verapamil
Cholesterol-lowering drugs: atorvastatin, cerivastatin, lovastatin (prodrug), simvastatin (prodrug)
Corticosteroids: hydrocortisone, methylprednisolone
HIV protease inhibitors: idinavir, nelfinivir, ritonavir, saquinavir
Hormonal agents: estradiol, progesterone
Immunosuppressant agents: cyclosporine, tacrolimus
Prokinetic agent: cisapride
Antibiotic agent: rifampin
Anticonvulsant agent: carbamazepine
Antidiabetic agent: insulin
Foods: chargrilled meats
Recreational drug: tobacco
Antibiotic agent: rifampin
Barbiturates: phenobarbital, secobarbital
Antibiotic agent: rifampin
Anticonvulsant agent: carbamazepine
Hormonal agent: norethindrone
Antibiotic agent: rifampin
Antibiotic agent: isoniazid
Recreational drug: ethanol, tobacco
Antibiotic agents: rifabutin, rifampin
Anticonvulsant agents: carbamazepine, phenytoin
Barbiturates : phenobarbital, secobarbital
Corticosteroids: dexamethasone, hydrocortisone, prednisolone, methylprednisolone
Herbal remedy: St. John’s Wort
HIV NNRTIsa: efavirenz, nevirapine
Oral hypoglycemic agents: pioglitazone, troglitazone
Antibiotic agent: ciprofloxacin, enoxacin, erythromycin, ofloxacin
Antidepressant agent: fluvoxamine
Antiplatelet agent: ticlopidine
H2 receptor blocker: cimetidine
Antibiotic agents: isoniazid, metronidazole, sulfamethoxazole, trimethoprim
Antidepressant agents: fluvoxamine, paroxetine, sertraline
Antidysrythmic agent: amiodarone
Antifungal agents: fluconazole, miconazole
Anticonvulsant agent: felbamate
Antidepressant agents: fluoxetine, fluvoxamine, paroxetine
Antiplatelet agent: ticlopidine
Proton pump inhibitors: lanzoprazole, omeprazole
Antidepressant agents: fluvoxamine, paroxetine, sertraline
Antidysrythmic agents: amiodarone, quinidine
H1 receptor blockers: chlorpheniramine, hydroxyzine, promethazine
H2 receptor blockers: cimetidine, ranitidine
NSAID (Nonsteroidal anti-inflammatory drug): celecoxib
Recreational drug: cocaine
Alcoholism rehabilitation drug: disulfiram
Antibiotic agents: ciprofloxacin, clarithromycin, erythromycin, norfloxacin
Antidepressant agents: fluvoxamine, nefazodone
Antidysrhythmic agent: amiodarone
Antifungal agents: fluconazole, itraconazole, ketoconazole
Calcium-channel blockers: diltiazem, verapamil
Food product: grapefruit juice
H2 receptor blockers: cimetidine
HIV protease inhibitors: idinavir, nelfinavir, ritonavir, saquinavir
The cytochrome P450 3A subfamily
The CYP3A isoforms account for up to 35% of total CYPs expressed in the liver and more than 80% of total CYPs expressed in the small intestine enterocytes (Fig. 2) [22, 23]. Intestinal CYP3A levels are generally 10–200% of those found in the liver and are 1.6-fold higher in women than in men . In addition, active CYP3A content in liver and small intestine is highly variable among individuals. For example, CYP3A4 protein varies up to 30-fold in S9 fractions prepared from 20 human duodenal biopsies and up to 100-fold in human liver microsomes (HLM) [22, 42]. The CYP3A subfamily is responsible for the metabolism of approximately 60% of clinically and toxicologically important agents, and to be inducible by steroids, macrolide antibiotics, imidazole antifungals and phenobarbital [19, 28, 39, 43–45]. Indeed, due to the anatomic arrangement of the small intestine and liver, drugs may encounter sequential, CYP3A-mediated first-pass metabolism when taken orally, and extensive liver metabolism when taken intravenously . The CYP3A subfamily in man appears to be composed of at least three genes, CYP3A4, 3A5 and 3A7. As CYP3A7 is only expressed in the fetal liver, accounting for 30–50% of total fetal CYPs, it will not be discussed in the current manuscript. CYP3A5 is polymorphically expressed in that it was observed in only 25% of adult Caucasian liver specimens examined, and in as much as 80% of the black African and African-American population [46, 47]. The expression of CYP3A5 does not appear to be influenced by gender or drug history, unlike CYP3A4. The metabolic capabilities of CYP3A5 have been shown to be very limited by comparison to CYP3A4, although the former isozyme has shown significant contribution to hepatic midazolam hydroxylation (97% of its elimination pathway), lidocaine demethylation, dextromethrophan N-demethylation and carbamazepine epoxidation in HLM and also in vivo [46, 47]. In fact, CYP3A5 does not appear to metabolize erythromycin, quinidine, terfenadine or 17α-ethynilestradiol at significant rates although testosterone, nifedipine, cortisol and dehydroepiandrosterone 3-sulfate were metabolized by CYP3A5, albeit at rates slower than CYP3A4 [28, 46]. Although the relative expression of the various 3A subfamily members in the human liver (i.e., the contribution of CYP3A5 to total CYP3A expression) would govern the rate and extent of the metabolism of the large number of compounds indicated earlier as substrates of CYP3A4, this discussion will mainly focus on CYP3A4-mediated metabolism and DDIs for clarity purpose.
As shown in Table 1, the substrates of CYP3A4 are structurally diverse and exhibit a wide range of sizes and affinities, some also show atypical kinetic profiles including positive cooperativity and substrate inhibition . In addition, several CYP3A4 specific inhibitors and inducers are shown in Tables 2 and 3 . The interactions between CYP3A4 and its substrates and inhibitors are thought to be complex, and may result in competitive or non-competitive inhibition, mechanism-based (irreversible) inhibition or activation depending on the combination investigated . It has been suggested that the complex effects observed with substrates of CYP3A4 are attributable to its spacious active site and its substrate-dependent CO-binding kinetics modulating enzyme conformation allowing the oxidation of large polycyclic aromatic hydrocarbons as well as the binding of multiple substrates within the active site of the enzyme [19, 34, 48, 49].
Phase II enzymes
Phase II enzymes are also referred to as “conjugating enzymes” and as detoxifying enzymes because they attach various water-soluble molecules to drugs (or other compounds) . The products generated by phase II enzymes are often considerably larger than the original drug and usually have any activity. Several important phase II enzymes are mainly expressed in the liver and small intestine and include uridine diphosphate-glucuronosyltransferase, sulfotransferases, acetyltransterases, catechol-O-methyltransferase and glutathione S-transferase . In contrast to phase I enzymes, phase II enzymes are generally present in the body in larger quantities and are less sensitive to small local changes. Hence, they are not normally involved in major clinically relevant DDIs, although genetic polymorphisms of acetylation and glucuronidation have recently been associated with drug toxicity . As with CYPs, human liver or intestinal microsomes, hepatocytes or intestinal cell cultures and human recombinant enzyme systems are used to determine the role of phase II enzymes in the metabolism of a given drug and in the study of DDIs .
Phase III enzymes: P-glycoprotein and related transporters
P-glycoprotein (ABCB1, P-gp) is the product of the MDR1 gene in humans and was first characterized as the ATP-dependent transporter responsible for efflux of chemotherapeutic agents from resistant cancer cells [22, 52]. P-gp is located within the brush border on the apical (luminal) surface of mature enterocytes and on the apical surface of hepatocytes. Examples of CYP3A4 and/or P-gp substrates with relatively narrow therapeutic index include terfenadine, simvastatin, lovastatin, felodipine, amiodarone and midazolam. Typically, if a drug undergoes significant (greater than 60%) CYP3A4 metabolism in the gut and the liver in addition to P-gp-mediated efflux, the likelihood of a significant DDI increases, especially for patients taking multiple CYP3A4 and/or P-gp substrates . The well-conserved family of transmembrane proteins also includes multidrug resistance related proteins MRP1–MRP6 but the clinical relevance of the latter transport proteins in DDIs in humans has yet to be determined [22, 23]. In vitro, P-gp function is measured as secretory transport back into the intestinal (in intestinal loops or cell systems) or biliary canalicular lumens (in human “sandwich” cultured hepatocytes) lumens. The best example of DDI reported to involve P-gp has been published with the cardiac glycoside digoxin [53, 54]. In Caco-2 cells, a human cell culture system overexpressing P-gp, the efflux of digoxin was inhibited by quinidine with an IC50 of 14 μM and by dipyridamole with an IC50 of 1.5 μM, suggesting a potent inhibitory potential on P-gp-mediated digoxin secretion, which has been further confirmed clinically [24, 53]. Interactions that occur at drug transporters, especially P-gp, are beginning to explain many non-metabolic DDIs, emphasizing the need for in vitro models that assess the respective role of metabolizing enzymes and transporters in drug biotransformation.
Human liver microsomes (HLM) and recombinant human CYPs as in vitro systems to assess drug metabolism and interactions
Because human liver samples and recombinant human enzymes are now readily available, in vitro systems are being used as screening tools to predict the potential for in vivo DDIs . One approach is to determine if human recombinant (also called cDNA-expressed) enzyme-transfected microsomes is capable of metabolizing the drug . A second approach is to determine the effect of specific chemical or antibody inhibitors on the drug’s metabolism in pooled HLM, expressing most of the liver phase I and II enzymes, but not the transporters . A third approach is to determine the rates of metabolism of the drug in a battery of HLM previously characterized in terms of metabolism of model substrates of the involved enzymes, also called “probes”. A perfect correlation between rates of metabolism of the drug and the probe across the entire battery of samples supports a central role for the enzyme in question. These experiments lead to the determination of in vitro kinetic parameters especially the rate of metabolite formation (Vmax) and the substrate binding affinity for the enzyme (Km). Combining the use of these in vitro systems along with the use of specific inhibitors and antibodies has been successful in predicting DDIs involving CYPs prior to studies in vivo [28, 56]. Generally, one probe substrate is used to characterize the inhibition of each of the various CYPs, except for CYP3A4, which requires a more complicated approach . Many in vitro probes for CYP3A activity have been established, including erythromycin, nifedipine, midazolam, diazepam, steroids, terfernadine and cyclosporine, most of which being clinically relevant to the intensivist’s daily activity . Unfortunately, the findings from any one in vivo probe fails to accurately predict the metabolism of all other CYP3A substrates. The reasons for the lack of correlation are likely to include extrahepatic metabolism as well as differences in the pharmacokinetic properties of the various probes. It has also been suggested that the complex effects observed with substrates of CYP3A4 are attributable to the binding of multiple substrates within the active site of the enzyme [19, 49]. Indeed, the active site of CYP3A4 is large enough to accommodate bulky molecules (e.g., erythromycin and cyclosporine) and it may be capable of binding more than one molecule. As a consequence, it is hazardous to extrapolate DDIs from in vitro to in vivo situation or from one CYP3A4 substrate to another. It is possible for a given medication to be a potent inhibitor of one substrate binding site but a weaker inhibitor of another substrate binding site, which is unpredictable in the clinical setting due to the lack of specific in vivo probe of a given CYP3A4 binding site. As a result, one concentration of a drug may result in significant inhibition of metabolism of CYP3A4 substrates, while having insignificant inhibition of other CYP3A4 substrates. For this reason, it has become customary to use at least two structurally diverse CYP3A4 probes for in vitro inhibition studies . In two recent studies using HLM and microsomes from cells containing cDNA expressed human CYP3A4, a 29-fold variation in the IC50 values were obtained with 27 inhibitor compounds tested with different substrates [58, 59]. These results indicate that the effect of a single modulator on the metabolism of a substrate is dependent on the identity of the latter, emphasizing the need for multiple CYP3A4 probes to be employed for the in vitro assessment of CYP3A4-mediated DDIs .
To be of optimal use, in vitro drug metabolism systems must accurately predict the metabolic fate and clearance of a drug in man and also the magnitude and likelihood of any clinically important DDIs [20, 22, 23, 49]. Clearly, a better understanding of the variability associated with CYP3A both in vitro and in vivo is required to improve the current capabilities for prediction with drugs metabolized by this enzyme. Furthermore, using HLM, there will always be a high likelihood that CYP3A4, due to its particular broad specificity, will play some role in the metabolism of the vast majority of drugs. Thus, even a relatively selective substrate for CYP2C9 or CYP2D6 may have its microsomal kinetics “contaminated” with CYP3A4 complexities, particularly when the latter is induced. The phenomenon of activation must be incorporated into the treatment of in vitro data when prediction of in vivo events is the aim of the study. This is not a trivial issue and heteroactivation is likely to be an important source of variability between individuals due to different dietary intakes, hormonal changes and pathological status (e.g., during septic shock), which will compound further the issue of variability in expression of these enzymes.
The rationale for using human material for in vitro metabolism studies is long established. A recent comparative study of the metabolism of three well known CYP3A4 substrates (i.e., dextromethorphan, midazolam and diazepam) in bacterial (Escherischia coli), yeast (Saccharomyces cerevisiae) and human B-lymphoblastoid cells expressing the enzyme as compared to a panel of human livers demonstrated, that the maximum rate of either substrate metabolism (Vmax) was generally 2- to 9-fold higher in human B-lymphoblastoid cells than the respective yeast, HLM and E. coli membrane preparation, resulting in higher intrinsic clearance . These results emphasize the use of HLM as the preferred material for drug metabolism studies.
In vitro cells systems as a potentially useful tool during the preclinical step of drug development
Primary human hepatocytes
The first utilizations of cultured human hepatocytes were published more than 15 years ago when studying the inhibition of CYP3A-mediated cyclosporine metabolism in human liver . These and other studies suggested that cultured human hepatocytes, usually collected as surgical waste or rejected donor livers and further isolated by an appropriate method, were capable of yielding new and clinically important information concerning potential DDIs due to CYP3A4 induction or inhibition . Investigators have been able to identify some drugs that are potent inhibitors of CYP3A4 (such as miconazole, erythromycin and troleandromycine) or inducers of CYP3A4 (such as sulfinpyrazole, phenylbutazone, and more recently rifampin) [20, 32, 35, 39, 43, 45].
The Caco-2 cells
Although most drugs given in ICU patients are intravenously administered, the intestinal mucosa still plays a substantial role in determining systemic exposure. Indeed, the small intestine is variously affected by pathological or drug-induced hemodynamic alterations (i.e., systemic changes in blood pressure and cardiac output due to sepsis or vasoactive drugs and inotropes). These changes may alter intestinal peristalsis, mucosal function and by extension intestinal metabolism [13, 65]. It is known that CYP3A exists not only in the liver but also in the gut, playing an important role in the first-pass metabolism after oral administration of its substrates (Fig. 2b) [22–24, 66, 67]. Using immunostaining techniques, it has been shown that the CYPs are located into the columnar absorptive epithelial cells of the villus and the mast cells in the small intestinal wall, just below the microvillus border, and that their content decreases from proximal to distal small intestine [29, 42]. The anatomy of the enterocyte may optimize the catalytic activity of its CYPs since lipophilic compounds must traverse a high-density zone of CYPs and transporters (see below) prior to entering the body, making presystemic first-pass as important as liver first-pass in investigating DDIs in vitro and by extent, in vivo.
A number of factors combine to complicate investigations of small intestinal metabolism in vivo and in vitro. These include the relatively short life of enterocytes, the marked effect of route of administration of inducing agents, the quantitative and qualitative variations in CYPs and transporters, especially CYP3A4 and P-gp along the length of the intestine, variations in CYPs between crypt and villous tips and in the procedures for preparation of enterocytes and their microsomes [29, 42, 68]. Besides, recent studies have revealed the overlapping substrate specificity of CYP3A4 and the efflux transporter P-gp . Some authors have even pointed out the possibility that the synergistic effects of CYP3A4-mediated metabolism and P-gp-mediated efflux in the gut epithelium may result in an unexpectedly high first-pass effect in the gut after oral administration [22, 23]. Thus, the inhibition or induction of CYP3A4 and/or P-gp caused by DDIs may affect the first-pass effect in the gut with clinically relevant consequences, as previously observed . Therefore, evaluation of DDIs before absorption into the portal vein using, preferentially human, intestinal absorption in vitro models such as the human intestinal Caco-2 cells system is expected to allow more precise prediction of in vivo DDIs.
The human colon carcinoma cell line Caco-2 grown in vitro under standard culture conditions in the absence of inducers of differentiation spontaneously exhibits signs of structural and functional differentiation and polarization . Transmission and scanning electron microscopy show that the entire upper side of the monolayer is covered with typical brush border microvilli. Caco-2 cells have proved to be useful as a model for studying intestinal permeability and drug transports. Recently, Schmiedlin-Ren et al. have found that 1α,25-dihydroxyvitamin D3 and all-trans-retinoic acid resulted in a dose- and duration-dependent increase in the expression of 4–5 pmoles of metabolically active CYP3A4 and P-gp but no detectable expression of CYP1A1 and CYP2D6 [71, 72]. This model has been recently used to evaluate the role of serum proteins, often altered in ICU patients, together with first-pass metabolism and transport in the intestinal permeability of saquinavir and midazolam [72, 73]. The major advantage of Caco-2 cells is that they are a human cell line and do not suffer from the interspecies differences in the morphological and physiological characteristics of the intestinal cells. Moreover, the method does not require the use of animals. However, limits to the use of this model to predict intestinal first-pass and its role in clinically relevant DDIs include the small amount of CYP3A4 expressed as compared to the human duodenal mucosa, the rather high level of P-gp expression due to the colonic origin of this cancer cell line, the very slow rate of transport, the lack of intestinal motility which may play a role in drug absorption, as well as the long preparation time of cells (2–4 weeks after cells seeding) and interlaboratory variability [42, 74]. Other in vitro methods to study drug absorption, such as the Ussing chamber or the rat everted gut sac technique showed variation between laboratories, due to the animal species used, the age, the method of sacrifice or anesthesia, the site of the gastrointestinal tract and the experimental conditions used. Hence, results obtained with these methods may not be easy to extrapolate to in vivo situations and may be hardly used by computer-assisted prediction systems.
Futures directions: the “computer-assisted” prediction system
Determination of metabolic properties of new chemical entities, especially the correct enzyme inhibition model and the correct concentration of inhibitor at the active site of the enzyme(s) of interest, is one of the most important steps during the drug development process, especially for drugs with nonlinear pharmacokinetics, since nonlinearity in bioavailability can lead to interindividual variability in plasma concentrations . It has been previously speculated that less than 20% of randomly in vitro investigated DDIs could be judged as potentially interacting, thus requiring further cautious investigations using human in vivo studies taking the therapeutic range, pharmacokinetics/pharmacodynamics characteristics, and severity of the adverse effects into consideration [55, 75].
Nowadays, in vitro methods are used for early estimation and prediction of in vivo metabolism as well as the risk for DDIs related to inhibition or induction of drug metabolic enzymes . Based on in vitro half-life (that is, the rate of disappearance of half of parent drug) and intrinsic clearance, in vivo pharmacokinetic parameters, such as bioavailability and in vivo half-life can be calculated. Due to the risk for metabolism-based DDIs between currently marketed drugs, inhibition and induction in vitro automated high-throughput screens together with in silico prediction software are used to estimate the risk for clinically significant DDIs. The uncertainty and inaccuracy of predicting the extent and duration of in vivo DDIs currently stems from a lack of definitive models by which to assess likely substrate and inhibitor or inducer concentrations at the active site of metabolism, as well as accurate mechanisms of inhibition (i.e., competitive, partial competitive, non-competitive, mixed-type reversible or mechanism-based irreversible), contribution of presystemic drug extraction and the effect of inhibitors on the processes involved [10, 11]. Ito and colleagues have attempted to quantitatively predict in vivo drug clearance and DDIs using the parameters for metabolism, binding and transport obtained from in vitro studies . More than a 10-fold difference was observed between in vivo and in vitro drug metabolic clearance for about one-fifth of the 29 metabolic reactions tested by the authors, including benzodiazepines, metoprolol, fentanyl, quinidine, tolbutamide and warfarin. One explanation for the larger in vivo metabolic clearance observed may be that the contribution of metabolism in tissues other than liver was not taken into account [22–24]. Another explanation was the incorrect assumption of a rapid equilibrium of drugs between blood and hepatocytes . Finally, the livers used for the estimation of in vitro and in vivo intrinsic clearance were completely different, so interindividual variability can affect the difference observed, essentially due to genetic polymorphism, smoking, alcohol consumption, pathological conditions or concurrent drug administration . Interestingly, the intrinsic clearance of phenytoin (55 ml/min) determined in vitro in primary human hepatocytes was comparable with that determined in vivo (62 ml/min), whereas the clearance based on HLM (17 ml/min) underestimated the in vivo intrinsic clearance .
In theory, the degree of change in systemic concentration of a drug caused by an interaction is determined by the route of administration, fraction of hepatic clearance in total clearance, fraction of the metabolic process subject to induction or inhibition, unbound concentration of the modulator around the enzyme and of the substrate . For instance, a 87% reduction in the clearance of triazolam when coadministered with the well-known CYP3A4 inhibitor ketoconazole at 200 mg/day was predicted taking all the above parameters into account, and was very close to the observed reduction in vivo .
As shown in previous works focusing on CYP3A4, the selection of a single substrate to investigate the effects of chemical modifiers on CYP3A4 should be approached with some caution. It is indeed possible that the extensive inhibition or induction of CYP3A4 caused by some drugs may under other experimental conditions be either overlooked or underestimated leading to errors in predicting important DDIs. One solution would be to study all potential in vivo combinations in vitro, focusing on the daily clinical practice. However, this would be very onerous and probably unnecessary. The current recommendation for investigators screening for inhibition or induction of CYP3A4 (but probably other CYPs)-mediated metabolism to highlight potential DDIs is to use at least three CYP3A4 assays. A recent database containing information about the clearance route for over 300 drugs from multiple therapeutic classes was constructed to assist in the semi-quantitative prediction of the magnitude of potential interactions with drugs under development . With knowledge of the in vitro inhibition constant of a drug (Ki) for a particular CYP isoform, it was theoretically possible to assess the likelihood of interactions for a drug cleared through CYP-mediated metabolism. For agents for which the CYP isoform involved in metabolism has not yet been identified, there is still substantial uncertainty given the current knowledge base.
To predict and help manage the occurrence of CYP-dependent DDIs, Bonnabry and colleagues recently developed an original computer application named Q-DIPS (standing for quantitative drug interactions prediction system) . The authors created their database to collect qualitative and quantitative data describing substrates, inhibitors and inducers of specific CYPs. Q-DIPS computer application gave up-to-date information, in dynamic tables, describing which specific CYP isozyme metabolizes a given drug, as well as which drugs may inhibit or induce a given isozyme . To better answer common clinical questions and help to rapidly evaluate the risk of interactions, it was possible to obtain an overview of substances causing DDIs with a specific drug or rather to focus on the patient’s prescription, thanks to a module allowing input with commercial names. For each question, key references, relevant quantitative data and quality indices were easily accessible. Based on enzymatic and pharmacokinetic data generated by in vitro and in vivo studies, the extrapolation module integrates quantitative models to predict the impact of a treatment on enzymatic activities. This new “computerized” approach showed promising potential for helping to improve DDIs management involving metabolism. Pending validation of extrapolation techniques used in the current computer application, in view of including factors such as intra-hepatocyte drug accumulation, physicians may indeed be able to anticipate the risk and extent of the identified DDI in a considered patient, which in turn may certainly help to individualize treatment decision and patient management. This should also help clinicians to focus on the characteristics of individuals who present with extreme reactions to therapy .
Drug interactions are one of the most dangerous and unrecognized complications of drug therapy, especially in ICU patients who are at an exceptionally high risk. As evidenced by both the large number of medications that are metabolized by and/or interact with metabolizing enzymes and the increasingly rapid rate at which new drugs are being approved, it is virtually impossible to predict and prevent all DDIs. However, by understanding the mechanism through which they occur, many DDIs can be prevented or detected early. In intensive care medicine, DDIs by mutual inhibition between drugs is almost inevitable, because CYP-mediated metabolism represents a major route of elimination of many drugs. The clinical significance of a metabolic DDI depends on the magnitude of the change in the concentration of active species (parent drug and/or active metabolites) at the site of pharmacological action and the therapeutic index of the drug. The smaller the difference between toxic and effective concentration, the greater the likelihood that a DDI will have potentially serious consequences. In addition, the often overlooked “clinical significance” of the observed DDI should always be considered in view of the clinicians’ experience, which is the goal of the combined “basic science-clinical implication” approach proposed in the current “mini-series”. The accurate prospective prediction of DDIs requires rigorous attention to the details of the in vitro results, and detailed information about the pharmacokinetics and metabolism of the substrate and modifier drugs. With the literature data concerning the clearance of various drugs, a framework for reasonable semi-quantitative predictions may be achieved. An effort to build virtual populations using extensive in vitro along with demographic, genomic, physiological and pathological data to simulate and predict drug disposition and potential for DDIs in patients in the ICU will probably be the next step to achieve optimal management of drug toxicity and interactions in clinical practice.