Clinical Studies of Drug-Drug Interactions:
`Design and Interpretation
`
`David J. Greenblatt and Lisa L. von Moltke
`
`Abstract The potential importance of drug-drug interaction (DDis) is increasing
`as polypharmacy becomes more and more prevalent. In vitro data cannot ~Erectly
`predict clinical DDis, but may provide a rationale for initiation of human stud(cid:173)
`ies to confirm or exclude possible interactions. Clinical DDI studies are designed
`to determine whether there is a real drug interaction not due to chance, how
`big the interaction is, and whether the DDI is of clinical importance. Statistical
`significance is not equivalent to clinical significance, and supplemental pharmaco(cid:173)
`dynamic or clinical outcome information is needed to address the importance of a
`pharmacokinetic DDI.
`·
`
`24.1 Introduction
`
`Drug-drug interactions (DDis) have become a topic of substantial scientific and
`public health concern over the last 20 years. While the clinical phenomenon of
`DDis had been recognized for a number of decades, several events in and around
`the years 1988-1993 brought the topic of DDis to a position of high attention
`and priority in the scientific community, as well as in the public arena. During
`this period, multiple human cytochrome P450 (CYP) isoforms became identi~ed,
`along with increasing understanding of their substrate and inhibitor specificities,
`relative quantitative importance in human drug metabolism, and mechanisms of
`genetic regulation (Clarke, 1998; Smith et al., 1998; b; Venkatakrishnan et al.,
`2001; Venkatakrishnan et al., 2003). Of particular importance in this context was
`CYP3A, with its unique hepatic and enteric distribution, and its major contribu(cid:173)
`tion to clearance of many clinically relevant drugs as well as naturally occurring
`chemicals (Venkatakrishnan et al., 2001; Venkatakrishnan et al., 2003; Guengerich,
`
`D.J. Greenblatt (1:81)
`Department of Pharmacology and Experimental Therapeutics, Tufts University School of Medicine
`and Tufts Medical Center, Boston, MA, USA
`e-mail: dj.greenblatt@tufts.edu
`
`(2010) . Clinical studies of drug-drug
`Greenblatt DJ and von Moltke LL
`interactions: design and interpretation. In, Enzyme-
`and Transporter(cid:173)
`Based Drug-Drug Interactions: Progress and Future Challenges. Edited by
`K. s. Pang, A. D. Rodrigues and R. M. Peter. New York, Springer: 625-
`649.
`
`1
`
`TEVA1025
`
`

`

`626
`
`D.J. Greenblatt and L.L. von Moltke
`
`1999; Gree~blatt et al., 2008). At the same time, in vitro techniques for studying
`human drug metabolism became increasingly developed and refined, including pre(cid:173)
`dictive models for in vitro-in vivo scaling, and the availability of heterologously
`expressed individual human CYPs. At a clinical level, polypharmacy was becoming
`increasingly prevalent, as the population aged, the number of patients with multi(cid:173)
`ple illnesses increased, and our capacity to provide pharmacologic treatments for
`serious disorders became more and more effective. Some newly introduced classes
`of medications - such as the azole antifungal agents and the selective serotonin
`reuptake inhibitor (SSRI) antidepressants - offered unique therapeutic options, but
`also had the secondary property of inhibiting certain human CYPs, thereby ele(cid:173)
`vating the risk of DDis (Greenblatt et al., 1999; Hemeryck and Belpaire, 2002;
`Venkatakrishnan et al., 2000). A dramatic and widely publicized event was the inter(cid:173)
`action of the nonsedating antihistamine terfenadine with potent CYP3A inhibitors
`such as ketoconazole and erythromycin (Honig et al., 1993b; Honig et al., 1992;
`Honig et al., 1994; Honig et al., 1993a). Under usual circumstances, terfenadine
`itself served only as a prodrug, being essentially completely transformed via hepatic
`and enteric CYP3A into fexofenadine, which was the entity having antihistaminic
`properties. Although terfenadine had effects on the cardiac QT c interval (Rampe et
`al., 1993; Crumb ~tal., 1995), this was of minimal concern since intact terfenadine
`does not ordinarily reach the systemic circulation. However, during co-treatment
`with CYP3A inhibitors, conversion of terfenadine to fexofenadine is blocked, and
`potentially hazardous levels of the parent drug reach the circulation (Honig et al.,
`
`60
`
`50
`
`CJ) z
`0
`~ 40
`0
`:::i
`ID
`:;:)
`fl.
`i5 30
`c
`L&.
`0
`ffi 20
`ID
`::::=
`:;:) z
`
`10
`
`•
`
`• • •
`
`•
`
`•
`
`\ ;
`
`•
`
`•
`
`•
`
`•
`
`•
`
`•
`
`•
`
`•
`
`•
`
`•
`
`•
`
`•
`
`•
`
`1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010
`YEAR
`
`Fig. 24.1 Number of articles indexed as DDI studies published per year in the Journal of Clinical
`Pharmacology, 1990-2008
`
`2
`
`

`

`24 Clinical Studies of Drug-Drug Interactions: Design and Interpretation
`
`627
`
`1994; von Mpltke et al., 1994b). A few cases of serious and even fatal cardiac
`arrhythmias ~ere reported as a consequence (Monahan et al., 1990; Woosley et al.,
`1993). The "terfenadine affair" led to an acutely increased awareness of the poten(cid:173)
`tial importance of DDis. Terfenadine was withdrawn from clinical practice, and a
`number of regulatory reforms increased the requirements for DDI assessments as
`a component of drug development. The overall shift in focus of the scientific and
`drug development community is clearly evident from the prevalence of DDI studies
`among scientific publications (Fig. 24.1).
`
`24.2 Epidemiology of Drug-Drug Interactions
`
`Given the prevalence of polypharmacy in contemporary clinical practice, the num(cid:173)
`ber of possible DDis can become very large. If an individual patient is taking n
`drugs concurrently, the number of pairwise combinations of these two drugs can be
`calculated as follows:
`
`n!
`(n- 2)!2!
`
`(24.1)
`
`The larger the value of n, the greater the number of different drug combi(cid:173)
`nation pairs, and potential pairwise DDis (Table 24.1). A patient with diabetes,
`hypertension, ischemic heart disease, and depression might well be taking 10
`drugs concurrently, in which case the number of possible drug interactions is 45.
`Considering this large "denominator" of possibilities, the number of clinically
`important DDis encountered in contemporary therapeutics actually is relatively
`small.
`)
`
`Table 24.1 Relation of
`number of drugs concurrently
`administered to the number of
`possible pairwise drug-drug
`interactions
`
`Number of drugs
`
`Possible pairwise drug interactions
`
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`
`1
`3
`6
`10
`15
`21
`28
`36
`45
`55
`66
`
`The outcome options following concurrent administration of two drugs can ·be
`constructed based on a probability hierarchy (Fig. 24.2). The most probable outcome
`is that the two drugs act independently, with no evidence of any interaction. Less
`probable is a DDI which can be demonstrated in a controlled laboratory setting,
`but is not detectable in clinical practice either because the magnitude of the change
`
`3
`
`

`

`628
`
`D.J. Greenblatt and L.L. von Moltke
`
`NO DDI
`
`"> zt: - ...
`en-elm
`II.ICI
`a:m uo
`lila: a a.
`
`DETECTABLE DDI,
`NOT CLINICALLY IMPORTANT
`~
`CLINICALLY IMPORTANT DDI,
`CAN BE SAFELY MANGAGED
`~
`USE THERAPEUTIC
`ALTERNATIVE
`FOR ONE OR BOTH DRUGS
`
`DRUG PAIR
`DANGEROUS,
`CONTRAINDICATED
`
`Fig. 24.2 Outcome possibilities in terms of DDis when two drugs are coadministered, in order of
`decreasing probability from top to bottom
`"")
`in plasma levels of the "victim" drug is so small as to be unimportant, or that the
`therapeutic index of the victim drug is very large. Still less probable is a DDI that
`is clinically important, but can be managed, for example, by reducing the dosage of
`the victim drug or by instituting closer monitoring of plasma levels and/or clinical
`outcome. Even lower in probability ranking is a DDI that is difficult to manage, such
`that co-treatment should be avoided if possible, and an alternative choice should be
`made for one or both drugs in the pair. The very least likely outcome - in fact, quite
`rare- is that the DDI potential carries an unacceptably serious risk, and the drug
`pair is contraindicated.
`This probability hierarchy has been confirmed in studies of the epidemiology
`of DDis. In a study of 9481 ambulatory patients in Germany, 13,672 actual drug
`combination pairs were identified (Bergk et al., 2004). Of these pairs, only 6.4%
`were known to cause DDls with predicted outcome of moderate or major severity,
`and 0.5% were unmanageable DDis such that the pair should be avoided. Findings
`were similar in a study of hospitalized patients in Denmark (Glintborg et al., 2005).
`The authors conclude that "although potential drug-drug interactions are highly
`prevalent, serious and clinically significant interactions are rare among recently
`hospitalized patients." In the specialty area of clinical psychopharmacology, there
`is extensive literature on the capacity of fluoxetine and paroxetine to inhibit the
`activity of human CYP2D6, leading to large inhibitory pharmacokinetic interactions
`
`4
`
`

`

`r
`
`24 Clinical Studies of Drug-Drug Interactions: Design and Interpretation
`
`629
`
`with CYP:?D6 substrate drugs such as desipramine (Hemeryck and Belpaire, 2002;
`von Moltk'e et al., 1994a; von Moltke et al., 1995; Preskorn et al., 1994; Alderman
`et al., 1997). Yet clinically important drug interactions are rarely reported in actual
`practice (Davies et al., 2004; deVane·, 2006; Molden et al., 2005). One possible
`explanation is that the therapeutic index of the victim drug or drugs is large enough
`that even a substantial change in plasma levels is not clinically relevant. Another
`explanation is that clinicians recognize the potential DDI, and make a pre-emptive
`downward adjustment in the dose of the victim to prevent the DDI.
`
`24.3 Drug Interaction Mechanisms and Terminology
`
`We have used the term "perpetrator" to indicate the drug that is causing the DDI,
`while "victim" or "substrate" is the drug that is being interacted with (Greenblatt
`and von Moltke, 2008). In a pure pharmacodynamic DDI, the perpetrator does
`not alter the plasma concentrations or systemic pharmacokinetics of the victim.
`Instead, the two drugs produce either additive or antagonistic pharmacodynamic
`effects. The interaction may occur via additive or opposite actions on the same
`receptor systems that yield additive or opposite clinical actions. Ethyl alcohol and
`benzodiazepines produce additive sedation through their actions on the gamma(cid:173)
`aminobutyric acid (GABA) receptor system (Chan, 1984; Greenblatt and von
`Moltke, 2008); the pharmacokinetic interaction between alcohol and benzodi(cid:173)
`azepines, if any, is small, and does not account for the additive sedative effects
`(Greenblatt et al., 1978; Greenblatt and von Moltke, 2008; Ochs et al., 1984;).
`Benzodiazepin~ agonists and caffeine have antagonistic pharmacodynamic actions.
`Benzodiazepines produce sedation via the GABA-benzodiazepine receptor sys(cid:173)
`tem, whereas caffeine produces alertness due to its action as an adenosine receptor
`antagonist (Biaggioni et al., 1991; Kaplan et al., 1992a, b; Kaplan et al., 1993).
`When caffeine is given together with a benzodiazepine agonist such as zolpidem,
`the sedative effects of zolpidem are partially reversed (Cysneiros et al., 2007).
`However, there is minimal, if any, pharmacokinetic interaction between these two
`agents.
`A pure pharmacokinetic interaction involves only the effect of the perpetrator
`on the systemic clearance of the victim drug, causing plasma levels of the victim
`to increase or decrease. The clinical actions of the victim may be correspondingly
`increased or decreased, but only because of the indirect effects of the perpetrator
`on systemic clearance, rather than a direct effect of the perpetrator on the target
`receptor mediating clinical action.
`Pharmacokinetic DDis involving drug-metabolizing enzyme systems (such as
`the CYPs) are generally classified as inhibition or induction. With metabolic inhibi(cid:173)
`tion, the perpetrator impairs the clearance of the victim drug, systemic exposure
`increases, and the clinical concern is toxicity. With induction, clearance of the
`victim increases, systemic exposure decreases, and the clinical concern is lack of
`efficacy (Table 24.2). However, inhibition and induction are not simply the same
`process in opposite directions- they involve fundamentally different mechanisms.
`
`5
`
`

`

`630
`
`D.J. Greenblatt and L.L. von Moltke
`
`·Inhibition of _Yictim drug clearance happens rapidly upon exposure to the perpe(cid:173)
`trator, and r~presents a11direct effect of the perpetrator on the drug-metabolizing
`enzyme. Metabolic inhibition can be studied in vitro using cell homogenates from
`human liver, or cells expressing human metabolic enzymes (Venkatakrishnan et al.,
`2001; Venkatakrishnan et al., 2003 ). From these in vitro systems it is straightforward
`to derive metrics of inhibitory potency such as the inhibition constant (Ki) or the
`50% inhibitory concentration (ICso). In contrast, induction is an indirect process(cid:173)
`the perpetrator (inducer) initiates a signal for the cell to produce more metabolic
`protein. This is slower than inhibition, and requires cultures of intact cells to study
`in vitro. The metric of induction potency is not so straightforward. Generally, the
`inductive effect of a candidate inducer is expressed as the fractional degree of induc(cid:173)
`tion relative to the hypothetical "maximum" induction by an index inducer such as
`rifampin.
`
`Table 24.2 Comparison of metabolic inhibition and induction
`
`Inhibition
`
`Induction
`
`Effect on victim drug
`Clearance
`Plasma levels
`Principal clinical concern
`Onset (after exposure to
`perpetrator)
`Offset (after perpetrator is
`discontinued)
`Mechanism
`
`In vitro system
`Metric of potency in vitro
`
`Decreased
`Increased
`Toxicity
`Rapid
`
`Rapid
`
`Increased
`Decreased
`Loss of efficacy
`Slow
`
`Slow
`
`Direct chemical
`effect
`Cell homogenates
`Ki or ICso
`
`Indirect signal to increase
`protein synthesis
`Cell culture
`Induction relative to maximum
`
`24.4 The Design of Clinical Drug Interaction Studies
`
`The general objective of DDI studies is to answer the following scientific questions:
`
`1. Given candidate "victim" and "perpetrator" drugs, is there a pharmacokinetic
`interaction between these two drugs that is not a chance event?
`2. What is the magnitude of the pharmacokinetic interaction?
`3. Is the interaction likely to be of clinical importance?
`
`Answers to the first two questions are largely objective and numerical, with little
`need for subjective interpretation or supplemental information. The third question is
`different- unless the DDI study incorporates measures of pharmacodynamic effect
`that are applicable to the target patient population, some supplemental information
`
`6
`
`

`

`24 Clinical Studies of Drug-Drug Interactions: Design and Interpretation
`
`631
`
`on the exp9sure-response relationship for the victim substrate drug is needed before
`a judgment can be ~de.
`
`24.4.1 Study Rationale
`
`The majority of clinical DDI studies involve healthy volunteers who do not have
`a medical need for the drugs under study. As such, study participation for these
`individuals is of no clinical benefit, but does entail some risk (though presumably
`low, and acceptable to an Institutional Review Board). There is also a dollar cost
`involved in the conduct of DDI studies. The cost is borne by the pharmaceutical
`sponsor in the case of an investigational drug, by the general public in the case of an
`NIH-supported study, or by some other entity. The core assumption is that the risk
`and cost of the DDI study are justified based on the potential public health benefit
`of the information to be acquired.
`Clinical observations raising the possibility of a DDI may form the basis for
`initiating a formal study to either confirm or rule out a DDI. In the course of drug
`development, in vitro data is commonly used to identify drug pairs for which a DDI
`needs to be evaluated in a clinical study. "Drug X" may be identified as an inhibitor
`of a certain human CYP isoform in vitro, with a quantitative potency metric of Ki or
`ICso. If [I] is a typical plasma concentration of Drug X encountered during treatment
`with the highest approved dosage, then the ratio [1]/Ki or [J]IICso is used to judge
`whether a clinical DDI is unlikely, possible, or probable, based on FDA guidelines.
`A DDI is termed "possible" if
`
`[l]/Ki > 0.1
`
`(24.2)
`
`This boundary is arguably too conservative on scientific grounds and triggers a
`large number of clinical DDI studies which tum out to be negative. Nonetheless that
`boundary reflects the current regulatory outlook, and sponsors often will initiate a
`DDI study on that basis.
`A second category of rationale for DDI studies is not directly scientific, but rather
`epidemiologic, based on a high probability of concurrent drug therapy. Drug X may
`be under development for a medical condition (such as diabetes, hypertension, or
`hyperlipidemia) that has high co-morbidity with ischemic heart disease. The sponsor
`may choose to initiate DDI studies of Drug X with digoxin or with warfarin because
`the probability of concurrent therapy is high, and because digoxin and warfarin (as
`potential victim drugs) have a narrow therapeutic index. Even if there is no direct
`scientific rationale raising the possibility of a DDI, it could be argued that clinical
`data excluding DDis with digoxin or warfarin is needed to assure safe co-treatment
`of Drug X with these potentially hazardous medications.
`Finally, an inevitable consequence of the industry-based system of drug devel(cid:173)
`opment is that research may be initiated solely for business reasons. Within a given
`
`7
`
`

`

`632
`
`D.J. Greenblatt and L.L. von Moltke
`
`drug class, a number of therapeutic options may be available, for which differences
`in therapeutic efficacy or toxicity may be subtle at most (Table 24.3). Competitive
`advantage then may turn on pharmacokinetic properties, such as mechanism of
`clearance, elimination half-life, and the risk of DDis. A clinical study may be initi(cid:173)
`ated to show that the sponsor's Drug X is not an inhibitor of a specific CYP isoform,
`whereas a competitor drug within the same class is in fact a significant inhibitor of
`the same CYP. These properties can be included in a product label, and used by
`pharmaceutical representatives or advertising materials for competitive advantage.
`An example is the interaction of macrolide antimicrobials with human CYP3A.
`Erythromycin, clarithromycin, and telithromycin are significant CYP3A inhibitors,
`whereas azithromycin is not (Greenblatt et al., 1998a).
`
`Table 24.3 Examples of
`drug classes for which
`individual drugs can be
`distinguished based on
`phannacokinetic properties or
`drug interaction potential
`
`Newer antidepressants
`Fluoxetine
`Sertraline
`Paroxetine
`Fluvoxamine
`Citalopram
`Venlafaxine
`Drugs to treat erectile dysfunction
`Sildenafil
`Tadalafil
`Vardenafil
`Macrolide antimicrobials
`Erythromycin
`Clarithromycin
`Azithromycin
`Telithromycin
`Hypnotics
`Triazolam
`Zolpidem
`Eszopiclone
`Temazepam
`
`24.4.2 Protocol Construction
`
`The customary design is a typical DDI protocol that involves a randomized, two(cid:173)
`way crossover study in a series of healthy volunteers. On one occasion, the victim
`substrate is administered in the control or baseline condition, without coadministra(cid:173)
`tion of the perpetrator. Total area under the plasma concentration curve from zero to
`infinity is calculated (AUC0). On a separate occasion, area under the curve is deter(cid:173)
`mined in the same subjects during coadministration of inhibitor (AUCJ). The AUC
`ratio (RAve) is calculated as
`
`AUC1
`RAve= AUCo
`
`(24.3)
`
`8
`
`

`

`24 Clinical Studies of Drug-Drug Interactions: Design and Interpretation
`
`633
`
`This represents the fractional increase in substrate AUC attributable to coad(cid:173)
`ministration of the perpetrator. The reciprocal of RAuc is the fractional change in
`clearance of the substrate.
`A key requirement is that the exposure of the volunteer subject to the perpetrator
`has to span the duration of blood sampling for plasma concentrations of the sub(cid:173)
`strate. The shorter the half-life of the substrate, the shorter the duration of sampling,
`and the lower the cost and risk of the D D I study. If the perpetrator drug is a metabolic
`inhibitor, this will prolong the necessary exposure duration and sampling time, but
`short half-life victim drugs, nonetheless, are easier to deal with in DDI studies. If
`the perpetrator is an inducer, this if anything decreases the necessary sampling dura(cid:173)
`tion, but this advantage may be offset by the need for a period of pretreatment with
`the inducer due to the time required to attain maximum induction.
`An alternative design is to study the kinetics of the victim drug at steady state.
`With this design, Equation (24.3) is modified to represent the ratio of substrate AUC
`values over a dosage interval segment at steady state. If the intrinsic kinetics of the
`victim drug are nonlinear, this may constitute support for the steady-state DDI study
`design. Beyond that, the steady-state design only has a "showcase" advantage in
`that it more closely mimics the usual therapeutic situation in which the substrate
`is given on an extended basis. However, if the kinetics of the substrate victim are
`linear (dose independent), single-dose kinetics are predictive of what will happen at
`steady state, and the single-dose design provides DDI data of equivalent quality. An
`obvious drawback of the steady-state design is that duration, cost, and risk of the
`study are substantially increased, since the substrate drug must be dosed to steady
`state both in the control condition and during coadministration of the perpetrator.
`
`24.4.3 Studies of Specific Drug Pairs
`
`The initiator of a DDI study may have a clinical or research question that applies
`only to a specific drug pair, without the objective of information that is more general(cid:173)
`izable. With the limited research objective, the study design involves administration
`of the substrate victim on two occasions, with and without perpetrator, as described
`above. The forthcoming research outcome applies to that drug pair, but not neces(cid:173)
`sarily applicable to any other pair. An example is the pharmacokinetic interaction of
`diazepam and fluvoxamine (Perucca et al., 1994), applicable to that particular com(cid:173)
`bination of substrate and perpetrator, but with no obvious connection to other drug
`combinations.
`
`24.4.4 Candidate Drug as Victim
`
`"Drug X" may be identified as a potential DDI victim either through in vitro data,
`clinical observations, or both. The in vitro model may have identified the one or
`
`9
`
`

`

`634
`
`D.J. Greenblatt and L.L. von Moltke
`
`more CYP isoforms responsible for clearance of Drug X. The commercial entity
`developing Drug X, or a group of academic investigators, then pose the question:
`what happens to the in vivo clearance of Drug X if one or more of the CYP iso(cid:173)
`forms responsible for clearance is either induced or inhibited by a perpetrator. This
`question may represent a critical point in drug development. The outcome could
`influence the drug's clinical safety profile, and the degree of restrictiveness of the
`product label if the drug is eventually approved. The DDI study outcome could even
`lead to discontinuation of the drug as a development candidate.
`The choice of perpetrator in the DDI study usually will be whatever produces
`the "worst case" - that is, the interaction of largest possible magnitude. The scien(cid:173)
`tific community and the FDA largely agree on what those specific perpetrators are,
`sometimes termed "index inhibitors" or "index inducers" (Table 24.4 ). Whatever the
`degree of inhibition or induction produced by the index compound, no other perpe(cid:173)
`trator will be any worse. Ketoconazole and ritonavir are typical choices of index
`inhibitor for studies of substrate victims metabolized by CYP3A (Lee et al., 2002;
`Tsunoda et al., 1999; Knox et al., 2008; Greenblatt et al., 2000). The sponsor or
`investigator may also wish to concurrently study a less potent perpetrator, in which
`case the DDI trial design would be modified to become a three-way crossover. For
`example, a candidate drug that is a CYP3A substrate may be studied with ketocona(cid:173)
`zole and erythromycin as perpetrators, representing strong and moderate CYP3A
`inhibitors, respectively.
`The impact of a DDI on the clearance of a victim drug is greatest when that
`drug is extensively metabolized, and a single CYP isoform mediates clearance.
`Candidate victim drugs metabolized mainly by CYP3A isoforms are a target of
`concern, since inhibition of CYP3A by a strong inhibitor such as ketoconazole or
`ritonavir may produce large values of RAve (Equation (24.3)) (Lee et al., 2002;
`Tsunoda et al., 1999; Knox et al., 2008; Greenblatt et al., 2000). Concern is aug(cid:173)
`mented when the substrate victim has high clearance, and undergoes significant
`presystemic extraction after oral dosage (Fig. 24.3).
`An important feature of study design is the optimal duration of pre-exposure to
`the perpetrator drug prior to administration of the substrate victim. To minimize
`study cost and risk, exposure duration should be the minimum necessary to pro(cid:173)
`duce maximum inhibition or induction. In the case of CYP3A inhibition studies,
`there is strong data to indicate that 24 h of pre-exposure to ketoconazole or ritonavir
`is sufficient to produce maximal inhibition (Fig. 24.4) (Stoch et al., 2009). For a
`time-dependent (mechanism-based) CYP3A inhibitor such as erythromycin or clar(cid:173)
`ithromycin, 48 h of pre-exposure is sufficient (Okudaira et al., 2007). On the other
`hand, if the perpetrator is an inducer (rifampin), a pre-treatment period of 5-7 days
`is needed for induction to become maximal (Ohnhaus et al., 1989; Lin, 2006).
`
`24.4.5 Candidate Drug as Perpetrator
`
`If the candidate drug is being evaluated as a possible perpetrator of DDls, the
`study design requires selection of an index substrate - that is, a substrate victim
`
`10
`
`

`

`r I I I
`
`24 Clinical Studies of Drug-Drug Interactions: Design and Interpretation
`
`635
`
`40
`
`35
`
`-..J
`1: -
`
`E 30
`c,
`
`::::5 25
`cr:
`..J
`0
`N 20
`cr: c
`!i
`cr: 15
`::::5
`C/) cr:
`..J 10
`Q.
`
`&--~
`I
`\
`I
`\
`t
`\
`I
`~
`I
`'-- '\
`I
`"-
`I
`I
`I
`+
`
`1
`
`CONTROL
`
`5
`
`/
`
`I RAuc = 33.4
`
`WITH RITONAVIR
`
`'t!r--A.....__
`
`/
`..................
`
`-.............._
`
`----...........................
`
`--
`
`..................
`
`--i'.
`
`0
`
`4
`
`8
`
`12
`
`16
`
`20
`
`24
`
`Fig. 24.3 Plasma concentrations of midazolam, an index substrate used to study DDis involving
`CYP3A, after a single 3-mg oral dose administered to a healthy volunteer on two occasions: once
`in the control condition (with no perpetrator coadministered), and again during coadministration of
`ritonavir (3 doses of 100 mg over 24 h), a strong inhibitor of CYP3A. RAuc is defined in Equation
`(24.3)
`
`whose clearance is established as mediated largely or entirely by a specific CYP
`isoform. Representative index substrates are shown in Table 24.4. As an exam(cid:173)
`ple, if the candidate perpetrator is a potential inhibitor of CYP3A, a DDI protocol
`could be structured as a two-way crossover study, with buspirone administered in
`the control condition, and again during coadministration of the candidate inhibitor.
`Investigators may wish to modify the study to become a three-way crossover, with
`an additional buspirone trial during co-treatment with the index inhibitor (ritonavir
`or ketoconazole) as a "positive control" to demonstrate maximum possible inhi(cid:173)
`bition. This provides valuable additional information on the degree of inhibition
`by the candidate perpetrator in relation to the maximal inhibition achievable in the
`experimental setting.
`A limitation of a DDI study using an index substrate is that the outcome cannot
`be directly extrapolated to other substrates cleared by the same CYP isoform. In
`general, strong inhibitors are "strong" and weakinhibitors are "weak" regardless of
`the substrate. However, the actual numerical RAve value for a given inhibitor will
`vary among victim substrates cleared by the same CYP isoform (Venkatakrishnan
`et al., 2001; Venkatakrishnan et al., 2003; Greenblatt et al., 2008; Ragueneau(cid:173)
`Majlessi et al., 2007; Galetin et al., 2005; Obach et al., 2005; Brown et al., 2006;
`Obach et al., 2006). Factors contributing to variability among substrates include
`
`11
`
`

`

`636
`
`D.J. Greenblatt and L.L. von Moltke
`
`J Clin Pharmacol 2009; 49:398-406
`
`300
`
`s::.
`><
`...J
`
`-... 250
`E g, 200 -0
`
`::::)
`<C
`:E 150
`<C
`...J
`0
`~
`0 100
`:e
`<C c:
`0
`
`...J
`
`50
`
`PRE
`
`1st day
`
`2nd day
`
`5th day
`
`KETOCONAZOLE (400 mg/day)
`
`Fig. 24.4 Mean (±SE) total AUC for midazolam after oral administration of midazolam on four
`occasions (Stoch et al., 2009). PRE represents the control condition, prior to co-treatment with
`ketoconazole. Midazolam was administered again at the beginning of days 1, 2, and 5 of co(cid:173)
`treatment with ketoconazole. For the day 1 study, midazolam was given with the first dose of
`ketoconazole; for the day 2 study, midazolam was given at the beginning of the second day, with
`24 h of midazolam pre-treatment. The results indicate that maximal inhibition - equivalent to
`that observed at steady state on day 5 - is attained on day 2, after 24 h of pre-treatment with
`ketoconazole
`
`the intrinsic clearance of the substrate, the fraction of total clearance attributed to
`the specific CYP isoform, the extent of presystemic extraction after oral dosage,
`and the relative contribution of enteric and hepatic metabolism to net presystemic
`extraction. A typical example is the substantially different effect of ketoconazole
`coadministration on the kinetics of triazolam and alprazolam (Greenblatt et al.,
`1998c ). Both drugs are structurally similar CYP3A substrates, but differ in the other
`features described above (Greenblatt et al., 2008; Greenblatt et al., 2002).
`A research design dilemma arises when the candidate drug is suspected as being
`an inhibitor of more than one CYP isoform. Individual DDI studies could be con(cid:173)
`ducted, each with a separate cohort of volunteer subjects, and each utilizing an index
`substrate corresponding to the specific CYP isoform. A second approach is to con(cid:173)
`duct a single DDI study using a substrate drug "cocktail" (Fuhr et al., 2007; Tanaka
`
`12
`
`

`

`r 1
`
`r
`
`24 Clinical Studies of Drug-Drug Interactions: Design and Interpretation
`
`637
`
`Table 24.4 Representative index substrates, inhibitors, and inducers applicable to the design of
`drug interaction studies*
`
`CYP isoform
`
`Index substrates
`
`Index inhibitors
`
`Index inducers
`
`CYP1A2
`
`Caffeine
`
`Flu voxamine
`
`CYP2B6
`CYP2C9
`CYP2C19
`CYP2D6
`
`CYP3A
`
`Bupropion, efavirenz
`Flurbiprofen
`Omeprazole
`Desipramine,
`dextromethorphan
`Midazolam, triazolam,
`buspirone
`
`Clopidogrel
`Fluconazole
`Fluvoxamine
`Quinidine, paroxetine
`
`[Cigarette
`smoking]
`Rifampin
`Rifampin
`Rifampin
`[None known]
`
`Ritonavir, ketoconazole Rifampin
`
`*Table entries are intended to be representative, not inclusive.
`
`et al., 2003; Zhu et al., 2001; Chainuvati et al., 2003; Blakey et al., 2004; Gurley
`et al., 2002; Chow et al., 2006; Zhou et al., 2004; Christensen et al., 2003). Instead
`of separate studies, volunteer subjects receive a mix of substrates concurrently, or in
`close temporal proximity, in a single study. Many possible substrate combinations
`have been proposed and utilized in cocktail DDI studies. A key piece of prelimi(cid:173)
`nary information is an unequivocal demonstration that each pairwise combination
`of substrates in the cocktail doe's not itself create DDis with each other.
`
`24.4.6 Approach to Analysis of Data
`
`If the clearance of a substrate drug is not dependent on a polymorphically regu(cid:173)
`lated process, the population distribution of AUC values following a fixed dose of
`that substrate will not be consistent with a normal distribution, but rather will have
`a positive skew. Generally the skewed pattern is consistent with a log-normal dis(cid:173)
`tribution (Fig. 24.5) (Greenblatt, 2008; Friedman et al., 1986; Lacey et al., 1997;
`Greenblatt et al., 1989). In any given DDI study, the number of AUC values is
`usually not sufficient for a stable characterization of the underlying statistical dis(cid:173)
`tribution. Nonetheless a log-normal distribution is generally assumed, based on
`experience with larger population studies.
`Going