メインコンテンツにスキップ
ホーム / コンテンツ / ブログ / The Next Horizons in Predicting Drug-drug Interactions

The Next Horizons in Predicting Drug-drug Interactions

Physiologically-based pharmacokinetic (PBPK) modeling has arrived in prime time. This quantitative mechanistic framework, combining physiology with drug information and clinical trial design, has become an integral part of drug discovery and development. PBPK has also gained currency within industry and regulatory agencies. Its applications are numerous, including simulation of pre-clinical, healthy volunteer and special population small and large molecule PK. If information relating to the drug’s formulation is available, insights into its behavior relating to absorption and the potential impact of food can be modeled in virtual human populations. The area that is gaining the most attention with regard to the capabilities of PBPK modeling, is for drug-drug interaction (DDI) prediction. DDIs can be caused by multiple mechanisms, some of which are better characterized than others. In this blog post, I will discuss the current regulatory and industry perspectives on using PBPK to predict DDIs as well as the knowledge gaps, particularly relating to transporter proteins, that must be bridged to fully realize the potential of this technology.

The molecular mechanisms behind DDIs

The most heavily studied DDIs are those mediated by alterations in metabolic enzymes, namely the cytochrome P450 (CYP) superfamily of enzymes. The drug that alters the activity of the metabolic enzyme is called the “perpetrator,” and the drug which is a substrate of that enzyme is termed the “victim.”

If drug A is metabolized by a CYP and drug B inhibits the CYP’s activity, the plasma concentration of drug A will be higher than expected and potentially cause toxicity. CYP inhibition can either be via a reversible (competitive) or irreversible (time-dependent) mechanism. If drug A is metabolized by a CYP and drug B induces or increases the enzyme’s activity, then the plasma concentrations of drug A will be lower than expected and may cause drug A to be ineffective.

Drug transporters also play a vital role in governing drug concentrations in the blood and in various organs including the liver, brain, intestine, lung and kidney. Transporters can move drugs into the tissues (increasing tissue drug levels) and also remove drugs (reducing tissue levels), dependent on the location and function of the transporter within the tissue. These proteins often modulate intestinal drug absorption, hepatic/renal elimination, and can enhance the effectiveness of cholesterol-lowering statin drugs. Yet, transporter protein DDIs (both competitive inhibition and induction) can cause loss of drug effectiveness and toxicity, which can also occur with enzyme-based DDIs.

Enzyme-mediated DDI prediction by PBPK: a regulatory and industry perspective

Regulatory Perspective: In a paper published last year in CPT Pharmacometrics and Systems Pharmacology, Wagner and his FDA colleagues provided the agency’s perspective on using PBPK in various clinical pharmacology applications, including predicting clinically relevant DDIs.

The agency has the greatest confidence in using PBPK when the investigational drug is a CYP3A or CYP2D6 substrate and the victim of an enzyme inhibitor. In fact, substrate/inhibitor models that are verified with clinical data can be used to simulate untested clinical scenarios and support drug labeling. The ability of PBPK models to predict the effect of a time-dependent inhibitor and of an enzyme inducer on investigational drugs is less well established. Predicting DDIs mediated by non-CYP pathways remains challenging. Further research into areas displaying predictive uncertainty will be required. These areas include quantifying these enzymes’ expression levels, enhancing the design and analysis of in vitro assays, and determining appropriate scaling factors to extrapolate from in vitro to in vivo.

Industry Perspective: Confirming the growing confidence in PBPK for regulatory applications, a group of modeling and simulation experts from ten global pharmaceutical companies published a position article on PBPK modeling last year. In this paper, the authors provided an industry perspective on using PBPK to predict DDIs. Their confidence in the validity of this approach varied, depending on the application.

DDIs caused by modulation of CYP enzymes have received the greatest amount of study. Thus, the industry scientists cited moderate to high confidence in using PBPK to quantify potential DDIs mediated by reversible CYP inhibition or induction, in which only a single mechanism is operating. To highlight the perceived limitations and to advance in vivo DDI predictions after extrapolation from in vitro assays, they proposed advisory statements applying rigor to the drug-related data incorporated into the model and to the lack of suitable in vivo data to enable model performance verification.

Other scenarios— CYP-mediated DDIs including time-dependent CYP inhibition, combinations of induction and reversible inhibition, modulation of non-CYP enzyme pathways, and intestinal metabolism— require verification using clinical examples.

Therefore, regulators and industry scientists have reasonable confidence in predicting certain enzyme-mediated DDIs using PBPK approaches. However, both establishments acknowledge that further verification is required to garner confidence in predicting other enzyme DDI scenarios using PBPK.

Transporter-mediated DDI prediction by PBPK: a regulatory and industry perspective

Both groups have less confidence in transporter-mediated DDI predictions using PBPK than for CYP-mediated enzyme predictions.

Regulatory Perspective: Currently, PBPK models need further development to predict transporter-mediated DDIs. Inadequate information regarding scaling factors and predicting intracellular drug concentrations makes performing in vitro in vivo extrapolation (IVIVE) premature. These models are further complicated by the need to consider the interplay between transporters and enzymes.

Industry Perspective: Like their FDA peers, the industry scientists cited less confidence relative to that of CYP3A- and CYP2D6-mediated DDIs when using PBPK to quantify the magnitude of DDIs caused by transporter-mediated mechanisms. Again, they attributed this lower confidence to knowledge gaps regarding deriving appropriate inhibition constants from in vitro assays and difficulty in predicting intracellular drug concentrations from uptake and efflux transporters.

The next horizons in using PBPK to predict DDIs

Regarding predicting transporter-mediated DDIs, the regulatory and industry position papers highlighted the knowledge gaps relating to transporter-based IVIVE scaling factors. These scaling factors are determined by quantifying transporter abundances in tissues and in vitro assay systems. Multiple laboratories generate in vivo organ transporter abundances, particularly for human liver transporters. This transporter data helps generate representative virtual human populations. The ratio of in vivo to in vitro transporter expression, or relative expression factor (REF), is required to predict transporter-mediated drug disposition and DDIs by IVIVE strategies.

To facilitate generating suitable scaling factors, transporter protein abundances and their inter-individual variability are quantified using LC-MS/MS proteomic techniques for mechanistic transporter IVIVE-PBPK. Many laboratories now generate absolute abundances with reported values for transporter isoforms ranging widely between studies. Scientists attempting to translate these data into IVIVE-PBPK strategies must determine whether these values represent population variability or are biased by the quantification methods used.

My colleagues and I addressed this question in a two-part study which has been accepted in Drug Metabolism and Disposition. We sought to address the differences in transporter abundance quantification between two laboratories and how these differences may translate when employed in IVIVE-PBPK.

Part I investigated the impact of quantifying the abundances of P-glycoprotein (P-gp) and Breast Cancer Resistance Protein (BCRP) in Caco-2 cells and human jejunum. We used the same samples in two different laboratories, in which each laboratory used a different technique to digest the cell membrane fractions and then quantify the protein abundances by different proteomic strategies using LC-MS/MS. The intestinal Relative Expression Factor (REFi) scalar commonly employed in transporter-meditated IVIVE-PBPK was generated for each protein from both laboratories.

Part II investigated whether the P-gp-REFi generated from our laboratory could be utilized with P-gp activity data from another laboratory to capture the observed Digoxin-rifampicin DDI when modeled in the Simcyp Simulator. We also investigated the impact of the differences in the BCRP-REFi generated by each laboratory on a drug’s theoretical BCRP disposition in the Simcyp Simulator, with particular reference to enterocyte concentrations.

In Part I, an extensive meta-analysis of the transporter proteomic literature was undertaken to describe the different approaches taken by multiple laboratories to quantify transporter abundances. In Part II, an extensive meta-analysis of the transporter expression (mRNA and protein) was performed to highlight the differences in P-gp- and BCRP-REFi’s generated by different laboratories using different techniques.

Results and conclusions of our study

In Part I, we found a systematically higher P-gp abundance in all samples quantified at the University of Manchester compared to Bertin Pharma. For BCRP, there was a systematically higher abundance in Caco-2 cell samples quantified by BPh compared to UoM, but not in human intestinal samples. Consequently, a similar intestinal REFi, based on distal jejunum and Caco-2 monolayer samples, between laboratories was found for P-gp. However, a 2-fold higher REFi was generated between the labs for BCRP. We postulated that the difference in P-gp abundance is based on the choice of standard (peptide) used for quantification. The reason for the difference in BCRP abundance in Caco-2 cells between labs has not been established.

In Part II, we found a wide range of REFi for P-gp or BCRP generated from the literature. The REFi for P-gp generated from the University of Manchester (from Part I) versus the 5-fold higher value currently used in the Simcyp Simulator yielded only modest differences in Cmax and AUC that were within the observed ranges for orally administered digoxin. However, the digoxin-rifampicin DDI could only be captured via IVIVE-PBPK using the P-gp kinetic data that was generated in the same Caco-2 cell systems as the 5-fold higher REFi for P-gp. Thus, transporter-specific kinetic estimates and REFi’s must be from the same in vitro system.

When assessing the 2-fold difference in REFi-BCRP, the higher REFi lead to a more distal absorption and delayed Tmax, highlighting how laboratory-specific differences in REFi generation can alter the IVIVE-PBPK outcome.

This work furthers our knowledge on the impact of methodology on transporter protein abundance quantification and how it affects transporter-mediated drug disposition and DDI predictions within IVIVE-PBPK models. Scaling factors such as REFi’s should be determined using proteomic methods specific to the laboratory with activity from the same in vitro system as the REFi was generated. If a laboratory requires generating a REF or similar scaling factor, but does not have the relevant human tissue, we provide an approach for generating these data in the last paragraph of Part II.

Due to variability in in vitro system characteristics between laboratories, caution is advised when using transporter activity data from one source and combining it with REFi from another laboratory when developing IVIVE-PBPK strategies.

Our current capability to incorporate absolute abundance scaling in the Simcyp Simulator is limited to the liver. Insufficient data prevents scaling via absolute abundances in other organs such as the intestine, brain and kidney. In addition, we must develop kinetic assays that can derive intrinsic transporter kinetic and inhibitory constants using mechanistic in vitro modeling technologies like the Simcyp In Vitro Analysis Toolkit (SIVA) for input in PBPK models. By filling these knowledge gaps, we can improve the capabilities of PBPK models to predict transporter drug disposition and DDIs.

Understanding the mechanisms of complex drug-drug interactions

To learn more about the role of transporters in drug disposition and toxicity, read this review I wrote with my Certara colleagues, Drs. Sibylle Neuhoff and Amin Rostami, “Absolute abundance and function of intestinal drug transporters: a prerequisite for fully mechanistic in vitro–in vivo extrapolation of oral drug absorption” in Biopharmaceutics and Drug Disposition. I also encourage you to register for our Best Practices and Transporters focused workshops to learn more about regulatory aspects of PBPK model submission.

PBPK modeling supports investigating the mechanisms of DDIs. Read this case study to learn how Certara scientists applied PBPK modeling to understand complex DDIs, both for drugs in development and following marketing approval.

筆者について

By: Matthew Harwood

Powered by Translations.com GlobalLink OneLink Software