In Vitro/In Vivo Correlation for Drug-Drug Interactions
Characterizing the potential for drug-drug interactions is critical to underwriting patient safety as new chemical entities proceed through the drug discovery and development pipeline. In vitro experiments to characterize the type and extent of interaction have been developed to inform chemical modifications early in discovery and to estimate the magnitude of potential interactions as drugs progress into the clinic. Regulatory guidance provides flow schemes based on a comprehensive understanding of drug disposition to enable decision-making as to whether particular clinical interaction studies need to be run and, if so, how they may be designed. Integration of information from in vitro, in vivo, and clinical sources provides the basis for drug labeling and the safe administration of drugs post-launch.
Drug-drug interactions (DDIs) occur when the dosing of a drug influences the pharmacokinetics (PK) or pharmacodynamics (PD) of a second drug. It has been estimated that 1–5% of hospital admissions may be due to DDIs (Day et al. 2017). Drugs such as mibefradil, terfenadine, and nefazodone (QTc prolongation) or bromfenac, alosetron, and cerivastatin (toxicity) have been removed from the market due to a high potential for DDIs (Wienkers and Heath 2005). Drug labels may contain a black box warning if concern over DDI potential is great enough. Factors influencing the likelihood and severity of DDIs may include age (the very young and aged are more susceptible), disease state, genetics, and polypharmacy, where the risk of DDIs increases dramatically when a patient is taking four or more medications simultaneously (Jacubeit et al. 1990). Due to the importance of DDIs in the drug discovery and development process, the US Food and Drug Administration (in vitro and clinical FDA Guidance for Industry 2017a,b) and the European Medicines Agency (EMA Guideline on the investigation of drug interactions 2012) have released and continually update regulatory guidance on the design and performance of in vitro and clinical DDI studies, as well as decision trees to guide decision-making as to whether particular DDI studies may be necessary as a new molecular entity (NME) proceeds through the drug discovery and development pipeline.
DDIs may be PK or PD based. PK is the study of what the body does to a drug. Typically, plasma or serum concentrations of a drug are measured as a surrogate for the concentration of drug at the site of pharmacological activity. This information is then visualized as a plasma concentration-time profile and quantified with PK parameters such as the area under the plasma concentration-time curve (AUC) or the maximum observed concentration of drug (C max) as determinants of overall drug exposure. PK-based drug interactions typically involve the inhibition or inactivation of enzyme activity, or the enhancement of enzyme expression, leading increases or decreases in drug exposure, respectively. PK-based DDIs also include phenomena such as loss of exposure due to increased gastrointestinal pH observed upon co-administration with an acid-reducing agent (Chung et al. 2015). PD is the effect of a drug on the body. PD-based DDIs occur when drugs influence each other’s pharmacological effects directly, such as the co-administration of two sedatives to potentiate activity.
Methods and Assumptions
Brief Primer on In Vitro Enzyme Kinetics
Types of DDIs
Inhibition types and characterization
Lineweaver-Burk (1/[S] vs. 1/ν)
S and I compete for E binding
K m ↑, V max↔
Intersect on y axis
I binding alters E architecture, reducing activity
K m ↔, V max↓
Intersect on x axis
I binds only to the ES complex
K m ↓, V max↓
I binds to E and ES
K m ↓ or ↑, V max↓
Intersect with x < 0 and y > 0
Multisite or atypical
Multiple ES complexes possible
Analogous to the observations with substrates, enzymes with large or flexible active sites may exhibit atypical or two-site inhibition profiles, where the substrate and the inhibitor occupy the enzyme active site simultaneously. This simultaneous occupancy of the active site may result in partial inhibition, where an inhibitor incompletely inhibits the turnover of a substrate, even as the inhibitor level is increased. Two-site models may also be applied to in vitro inhibition data where fitting of competitive, noncompetitive, uncompetitive, and mixed inhibition models to observed data is poor. Two-site inhibition is one potential explanation for substrate-dependent inhibition, where the same inhibitor exhibits different apparent inhibition potency for the same enzyme when compared using different substrates (VandenBrink et al. 2012).
Mechanisms of CYP inactivation with diagnostic experiments
Loss of activity
Loss of CO binding
Loss of native heme
Formation of GSH, cysteine, lysine, or cyanide adducts
Gemfibrozil glucuronide (CYP2C8)
Appearance of peak at 440–450 nm (UV–vis spectrum); Fe(CN)6 restores activity
These parameters may be used to quantitatively estimate induction potential in the clinical situation. As the induction assay requires fully functional hepatocytes, a commonly encountered confounding factor in the determination of induction parameters is cell toxicity; toxicity is commonly observed with increasing concentration for cytotoxic drugs and may prevent full characterization of induction potential.
The presence of inflammatory cytokines may lead to a process called downregulation, which has the opposite effect of induction. Downregulation is most commonly seen with inflammatory diseases, where IL-6 and other cytokines may lead to an apparent reduction in transcription and expression of DMEs, reducing the turnover of DME substrates and leading to higher drug levels (Evers et al. 2013). Treatment of the inflammatory disease may reduce cytokine-mediated influence on DMEs, leading to increased transcription, protein synthesis, and DME turnover, leading to lower drug levels. Due to the complexities in the downregulation process, neither in vitro nor preclinical models may serve as predictors of the clinical situation with the current state of knowledge (Dickmann et al. 2011).
Experimental Considerations in Vitro
Typical CYP-selective probe substrates, inhibitors, and inactivators
K m (μM)
K i (μM)
K I (μM)
K inact (min−1)
PhRMA has also published an overview on the conduct and design of experiments to characterize TDI (Grimm et al. 2009). For TDI experiments, the additional experimental factors of concern are protein concentration in the preincubation, preincubation time, incubation time, dilution factor, and concentrations of the inactivator and probe substrate. Protein concentration in the preincubation is important because a balance of turnover versus nonspecific binding is desired; increasing protein will not necessarily translate to an increase in turnover or inactivation rate. Preincubation time is important, as it is generally desired to use initial rates for the inactivation calculations; too little or too much inactivation may increase the difficulty in calculating these kinetic parameters. For IC50 shift experiments, dilution from a preincubation is not done, so that the substrate concentration remains the same (at k m ) for the incubations with and without preincubation. Dilution is normally carried out for the K i and k inact experiment, as it is desired to minimize the potential for reversible inhibition ; the dilution also reduces the potential for inactivation during the probe substrate reaction. High substrate concentrations are used (often five- to tenfold of k m ), to ensure that sufficient substrate is present to outcompete remaining inactivator from the enzyme active site.
For in vitro induction experiments, key features identified by PhRMA include the exposure time of the hepatocytes to the NME, the concentrations of the NME used, the potential of the NME to cause cellular toxicity, and the experimental output to measure (Chu et al. 2009). Regulatory guidance from EMA and FDA both recommend the use of three or more hepatocyte donors; the hepatocytes may be cryopreserved or fresh. Changes in mRNA levels versus positive controls are the assay readout for development studies; DME activity may be assessed in some cases as a secondary endpoint. Regulatory guidance from EMA guidance recommends exposure to test article over 72 h. Both FDA and EMA guidance recommend refreshing exposure to the NME daily; EMA recommends measuring the amount of parent drug from the incubation at several times on the final day of the experiment. Single concentrations of NME may be used to characterize induction potential in the discovery environment, while multiple NME concentrations are used to characterize EC 50 and E max for the development environment. EMA guidance recommends in vitro concentrations exceeding 50-fold the mean unbound C max value for DMEs present in the liver, with in vitro concentrations of NME exceeding 0.1*dose/250 mL as an estimate for DMEs present in the gut (CYP3A). As the in vitro concentration of NME increases, the likelihood of cell toxicity may increase. Simultaneous loss of cell viability and reduction in mRNA levels are expected if cell viability alone is the cause; a reduction in mRNA levels without loss of cell viability may indicate downregulation of the DME by the NME.
Prediction of DDI for the Clinical Situation
This equation holds true for both high and low extraction ratio drugs (Benet and Hoener 2002).
Quantitative Prediction of Clinical DDIs
Regulatory Guidance and DDIs
An additional option is PBPK modeling. PBPK is a modeling technology that has seen recent emergence both for internal decision-making for compound progression through drug discovery and development and for regulatory applications (Jamei et al. 2009). PBPK models have three main types of input parameters: demographic and genetic information (i.e., age or gender), physiological information (i.e., organ blood flow and enzyme levels), and drug-specific parameters (i.e., pKa, logP, solubility). Differential equations and Monte Carlo-based simulations integrate the inputs together to simulate a variety of outcomes, including plasma concentration-time profiles, enzyme activity profiles, and drug-tissue concentrations. Because of the types of inputs and the modeling technique used, PBPK is well-suited for modeling where changes in physiology or populations may impact PK, changes in physicochemical properties or formulations may impact PK, or where dynamic simulations such as DDIs are desired. PBPK may be used to answer fundamental clinical pharmacology based questions such as (1) what are the intrinsic factors that may influence exposure, (2) what are the extrinsic factors that may influence exposure, and (3) what are situations in which dosing may need to be adjusted due to intrinsic and extrinsic factors. Because physiologically relevant parameters are included, PBPK may more closely represent the clinical situation than basic or MSM models. PBPK may also provide information on expected variability in studies based on demographic factors.
The potency of an inhibitor or inducer is determined based on the magnitude of its interaction with a sensitive probe substrate for a specific enzyme pathway. Strong inhibitors increase AUC ≥ fivefold, moderate inhibitors increase AUC two- ≥ to < fivefold, and weak inhibitors increase AUC ≥ 1.25- to < twofold. Strong inducers reduce AUC ≥ 80%, moderate inducers reduce AUC by ≥50 to <80%, and weak inducers reduce AUC by ≥20 to <50%.
Free Drug Hypothesis
The free drug hypothesis is a tenet of drug discovery and development, which has two fundamental premises (Smith et al. 2010) First, at steady state, drug concentration on either side of a biological membrane is the same. Second, it is free drug at the site of action that is pharmacologically active. While these two proposals may often be true, there are a numbers of situations where the free drug hypothesis may fail. The first premise may fail when drug permeability is low, when uptake transporters increase the concentration of a drug in a tissue, when efflux transporters decrease the concentration of drug in a tissue, or when low or disrupted blood flow may reduce the concentration of drug throughout a tissue. The second premise of the free drug hypothesis may fail when an irreversible inhibitor is involved, when a series of target-mediated events must occur in a series in order to elicit a pharmacological effect, or when in vitro conditions poorly represent the in vivo condition.
In an idealized situation, the ratio of in vitro K i to K i,iv should equal one. Marked deviation from unity indicates the presence of a situation not accounted for in the in vitro experiment, such as mechanism(s) other than reversible inhibition, an active site environment in the in vivo situation that differs from the in vitro conditions, or active uptake or efflux that has altered the relative concentration of perpetrator drug at the enzyme active site. This type of phenomena has been observed clinically for drugs such as fluvoxamine, which has a high liver to plasma partition ratio. Using (S)-mephenytoin as a probe substrate for CYP2C19 activity, the unbound K i,iv was estimated to be 1.9 nM, compared to an unbound in vitro K i of 76 nM, a nearly 40-fold increase in estimated inhibition potency in vivo (Yao et al. 2003). Similarly, the K i,iv for the fluvoxamine for its interaction with theophylline (CYP1A2) was estimated at 3.6 nM, while the in vitro K i was determined to be 36 nM. Conversely, the interaction of the reversible inhibitor fluconazole with the CYP2C9 substrate (S)-warfarin has an estimated K i,iv of 19.8 μM, similar to the measured in vitro K i of 8 μM, indicating similarity between the in vitro and in vivo environments (Neal et al. 2003).
Is Concern over Plasma Protein Displacement DDIs Justified?
As can be seen, f u is not involved in the final equation, and thus, changes in f u are not expected to impact orally administered drugs cleared by the liver. Two unlikely instances where plasma protein displacement could conceivably play a role in DDIs are cases where a narrow therapeutic index drug is dosed intravenously and exhibits a high extraction ratio or where a narrow therapeutic index drug is dosed orally and exhibits a very fast PK/PD equilibrium (Benet and Hoener 2002).
Impact of Pharmacogenetics on DDIs
Phenotype may have a marked role in DDIs in certain circumstances, particularly when the main route of clearance is due to a polymorphically expressed enzyme. If, for example, a NCE is primarily cleared by CYP2D6, secondary DMEs may take on the primary role in clearance for CYP2D6 PMs. Co-administration of a drug that inhibits the secondary pathway would normally have minimal effect on the NCE clearance for a CYP2D6 EM but may exhibit a marked DDI effect for the CYP2D6 PM. In vitro characterization of routes of clearance and phenotype assessment of patient characteristics can be pivotal for informing and preventing these types of DDIs.
The theory and practice of understanding and providing quantitative estimates for metabolism-based DDIs has advanced dramatically. In vitro assay design and screening provide information on chemotypes that likely exhibit DDIs in the clinical situation and provide medicinal chemists with the basis to minimize DDI impact through the development of structure-activity relationships early in the discovery process. With the advent of technologies like PBPK, drug metabolism and clinical pharmacology scientists may quantitatively predict the clinical effects of DDIs based on high-content in vitro experiments, even for complex situations where inhibition, inactivation, and induction are all expected simultaneously or where multiple drug entities such as drug metabolites are present. Current frontiers in the field include the prediction of DDIs based on drug metabolism-transporter interplay and the prediction of DDIs for special populations including pediatrics and specific disease states. Understanding of the DDI potential for a clinical drug candidate is gained through the careful generation, examination, and integration of in vitro, preclinical, and clinical data to form a cohesive picture of the absorption, distribution, metabolism, and excretion (ADME) characteristics for that drug.
References and Further Reading
- Bjornsson TD, Callaghan JT, Einolf HJ, Fisher V, Gan L, Grimm S, Kao J, King SP, Miwa G, Ni L, Kumar G, McLeod J, Obach SR, Roberts S, Roe A, Shah A, Snikeris F, Sullivan JT, Tweedie D, Vega JM, Walsh J, Wrighton SA (2003) The conduct of in vitro and in vivo drug-drug interaction studies: a PhRMA perspective. J Clin Pharmacol 43:443–469CrossRefPubMedGoogle Scholar
- Bohnert T, Patel A, Templeton I, Chen Y, Lu C, Lai G, Leung L, Tse S, Einolf H, Wang Y-H, Sinz M, Stearns R, Walsky R, Geng W, Sudsakorn S, Moore D, He L, Wahlstrom J, Keims J, Narayanan R, Lang D, Yang Q (2016) Evaluation of a new molecular entity as a victim of metabolic drug-drug interactions – an industry perspective. Drug Metab Dispos 44:1399–1423CrossRefPubMedGoogle Scholar
- Chu V, Einolf HJ, Evers R, Kumar G, Moore D, Ripp S, Silva J, Sinha V, Sinz M, Skerjanec A (2009) In vitro and in vivo induction of cytochrome P450: a survey of the current practices and recommendations: a pharmaceutical research and manufacturers of America perspective. Drug Metab Dispos 37:1339–1354CrossRefPubMedGoogle Scholar
- Chung J, Alvarez-Nunez F, Chow V, Daurio D, David J, Dodds M, Emery M, Litweiler K, Paccaly A, Peng J, Rock B, Wienkers L, Yang C, Yu Z, Wahlstrom J (2015) Utilizing physiologically based pharmacokinetic modeling to inform formulation and clinical development for a compound with pH-dependent solubility. J Pharm Sci 104:1522–1532CrossRefPubMedGoogle Scholar
- European Medicine Agency (EMA), Committee for Human Medical Products (CHMP) (2012) Guideline on the investigation of drug interactions. http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2012/07/WC500129606.pdf
- European Medicine Agency (EMA), Committee for Human Medical Products (CHMP) (2015) Guideline on key aspects for the use of pharmacogenomics in the pharmacovigilance of medicinal products. http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2015/11/WC500196800.pdf
- Evers R, Dallas S, Dickmann LJ, Fahmi OA, Kenney JR, Kraynov E, Nguyen T, Patel AH, Slatter JG, Zhang L (2013) Critical review of preclinical approaches to investigate cytochrome P450-mediated therapeutic protein drug-drug interactions and recommendations for best practices: a white paper. Drug Metab Dispos 41:1593–1609CrossRefGoogle Scholar
- FDA Guidance for Industry (2013) Clinical pharmacogenomics: premarket evaluation in early-phase clinical studies and recommendations for labeling. https://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM337169.pdf
- FDA Guidance for Industry (2017a) Clinical drug interaction studies – study design, data analysis, and clinical implications. https://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM292362.pdf
- FDA Guidance for Industry (2017b) In vitro metabolism and transporter-mediated drug-drug interaction studies. https://www.fda.gov/downloads/drugs/guidancecomplianceregulatoryinformation/guidances/ucm581965.pdf
- Grimm SW, Einolf HJ, Hall SD, He K, Lim HK, Ling KH, Lu C, Nomeir AA, Seibert E, Skordos KW, Tonn GR, VanHorn R, Wang RW, Wong YN, Yang TJ (2009) The conduct of in vitro studies to address time-dependent inhibition of drug-metabolizing enzymes: a perspective of the pharmaceutical research and manufacturers of America. Drug Metab Dispos 37:1355–1370CrossRefPubMedGoogle Scholar
- Korzekwa KR, Krishnamachary N, Shou M, Ogai A, Parise RA, Rettie AE, Gonzales FJ, Tracy TS (1998) Evaluation of atypical cytochrome P450 kinetics with two-substrate models: evidence that multiple substrates can simultaneously bind to cytochrome P450 active site. Biochemistry 37:4137–4147CrossRefPubMedGoogle Scholar
- Michaelis L, Menten ML (1913) Die kinetic der invertinwirkung. Biochem Z 49:333–369Google Scholar