Regulatory utility of physiologically‐based pharmacokinetic modeling to support alternative bioequivalence approaches and risk assessment: A workshop summary report

Abstract

This report summarizes the proceedings for day 2 sessions 1 and 3 of the 2‐day public workshop entitled “Regulatory Utility of Mechanistic Modeling to Support Alternative Bioequivalence Approaches,” a jointly sponsored workshop by the US Food and Drug Administration (FDA) and the Center for Research on Complex Generics (CRCG). The aims of this workshop were: (1) to discuss how mechanistic modeling, including physiologically‐based pharmacokinetic (PBPK) modeling and simulation, can support product development, and regulatory submissions; (2) to share the current state of mechanistic modeling for bioequivalence (BE) assessment through case studies; (3) to establish a consensus on best practices for using PBPK modeling for BE assessment to help drive further investment by the generic drug industry into mechanistic modeling and simulation; and (4) to introduce the concept of a Model Master File to improve model‐sharing. The theme of day 2 covered PBPK absorption model for oral products as an alternative BE approach and a tool for supporting risk assessment and biowaiver (session 1), oral PBPK for evaluating the impact of food on BE (session 2), successful cases, and challenges for oral PBPK (session 3). This report summarizes the topics of the presentations of day 2 sessions 1 and session 3 from FDA, academia, and pharmaceutical industry, including the current status of oral PBPK, case examples as well as the challenges and opportunities in this area. In addition, panel discussions on the utility of oral PBPK in both new drugs and generic drugs from regulatory and industry perspective are also summarized.

INTRODUCTION

On September 30 and October 1, 2021, the US Food and Drug Administration (FDA) and the Center for Research on Complex Generics (CRCG), a joint collaboration between the University of Maryland and the University of Michigan, hosted a virtual public workshop entitled “Regulatory Utility of Mechanistic Modeling to Support Alternative Bioequivalence Approaches.” ¹ The aims of this 2‐day workshop were to engage the generic pharmaceutical industry and other involved stakeholders regarding the following:

• How mechanistic modeling (including physiologically‐based pharmacokinetic [PBPK] modeling) and simulation can support drug development and regulatory submissions;

• Share the current state of mechanistic modeling for bioequivalence (BE) assessment through case studies;

• Establish a consensus on best practices for using PBPK and computational fluid dynamics modeling for BE assessment to help drive further investment into mechanistic modeling and simulation; and

• Introduce the concept of a Model Master File (MMF) to improve model‐sharing among model developers, industry, and the FDA.

Day 1 of this workshop explored mechanistic modeling for orally inhaled, dermal, and locally acting generic drug product development. ¹ The theme of day 2 focused on the use of PBPK absorption model for oral products (indicated as oral PBPK hereafter) as an alternative BE approach and a tool for supporting risk assessment and biowaiver (session 1), oral PBPK for evaluating the impact of food on BE (session 2), successful cases and challenges for oral PBPK (session 3), and model acceptance and model sharing for regulatory use (session 4). ¹

This workshop report provides a summary of presentations and extensive discussion points from day 2 sessions 1 and 3 of the FDA‐CRCG PBPK workshop. The scope of the other sessions is not addressed in this report. Day 2 session 1 and session 3 speakers from the FDA, academia, and the pharmaceutical industry presented the current status of oral PBPK, case examples as well as the challenges and opportunities in this area. These presentations were followed by panel discussion sessions focusing on the utility of oral PBPK in both new and generic drug development from regulatory and industry perspective. The panel discussion sessions included moderators and representatives from industry, academia, and the FDA.

PRESENTATIONS FOR DAY 2 SESSION 1

PBPK modeling has the capacity of integrating in vitro information (e.g., in vitro dissolution and particle size distribution) with physicochemical properties of drug substance and physiological factors to predict the systemic exposure to drug substance. Recently, the Agency published a draft guidance for industry ² outlining general approaches for incorporation of these factors into a PBPK model. Speakers from session 1 provided case examples on using oral PBPK as an alternative BE approach and a tool for supporting risk assessment, biowaiver, and BE assessment (e.g., evaluating the impact of changes in critical material/quality attributes on BE, setting dissolution safe space, assessing pH dependent drug–drug interaction [DDI] potential). A safe space is defined by the boundaries demarcated by in vitro specifications (i.e., dissolution or, when applicable, other relevant drug product quality attributes), within which drug product variants are anticipated to be bioequivalent to one another. Building a safe space may also reduce the need for in vivo data to support regulatory assessment. Although safe spaces can be used for both new and generic drug products, building a safe space for a generic drug product necessitates the identification of a range of dissolution profiles within which the proposed drug products are found to be bioequivalent to one another and to the reference or target drug product (e.g., via virtual BE analysis). ² Further, PBPK modeling for biopharmaceutics applications is a promising tool to promote the Patient‐Centric Quality Standards (PCQS) based on risk assessment and setting clinically relevant specifications. The presenters also discussed the challenges for implementing applications of PBPK modeling for BE assessment and some of the examples are as follows: lack of sufficient model validation, lack of using dissolution and solubility data that are biorelevant/biopredictive, subjective simulation findings depending on the modeler, and challenges in developing PBPK model under fed conditions to describe the interaction among food, gastrointestinal (GI) physiology (e.g., GI motility), and drug substance.

Summary of Presentation “PBPK Absorption Modeling to Support Risk Assessment and Biowaiver for Generic Oral Products” by Fang Wu, Senior Pharmacologist and Scientific Lead from Office of Research and Standards (ORS), Office of Generic Drugs (OGD), FDA.

Briefly, Dr. Fang Wu presented on using oral PBPK as an alternative BE approach and a tool for supporting risk assessment and biowaiver. ³ To facilitate the use of PBPK in regulatory submissions, the FDA issued two guidances in 2020 that describe the utility of PBPK absorption modeling for biopharmaceutics applications ² and predicting the pH‐dependent DDI potential. ⁴ In recent years, the role of PBPK modeling in regulatory submissions is expanding. Wu summarized the regulatory questions that the PBPK absorption model can address in generic drug development for BE assessment, which includes setting dissolution safe space, evaluating the impact of changes in critical quality attributes, assessing the risk of release mechanism change, pH dependent DDI potential, predicting GI local concentration, supporting BE assessment in specific populations, and waiving of in vivo studies. In addition, Wu presented a few case studies of oral PBPK applications obtained from the regulatory submissions. ³ One case was using PBPK modeling to assess the impact of particle size distribution (PSD) on BE, which showed a low risk of non‐BE when D90 changed over a wide range at fixed D50. Note that D90 means that the portion of particles with diameters below this value is 90% and D50 means that the portion of particles with diameters below this value is 50%. These modeling results supported a satisfactory BE assessment of this abbreviated new drug application (ANDA) and setting a clinically relevant three tier PSD specification. Another case example was on using PBPK/physiologically based biopharmaceutics modeling (PBBM) to predict the impact of faster release of lower strength of a generic drug product (drug X tablets) on the BE under fasting and fed conditions. The FDA identified several deficiencies in the submitted model, especially on assessing model performance in the proposed context of use, which impeded the use of the developed model in assessing the impact of dissolution differences on in vivo performance/BE. Wu also presented the challenges and research opportunities when developing PBPK models for supporting risk assessment and biowaiver. One of the challenge examples is the necessity of having a biopredictive dissolution method to establish a relationship between in vitro characteristics and observed or estimated PK. Other challenges include, but are not limited to: the modeling of in vitro dissolution profiles, the combination of drug‐ and formulation‐specific properties with GI variability, and the impact of the formulation/excipients on permeation for Biopharmaceutics Classification System (BCS) Class III drugs. Further, Wu clarified that the modelers need to inform the models with appropriate estimates of subject variability, which can be obtained from a previously conducted in vivo pharmacokinetic (PK) study or included in the system parameters. Considering the above challenges, research opportunities related to oral PBPK modeling have been conducted through both internal and external projects and resulted in several publications. ⁵ , ⁶ , ⁷ , ⁸ She concluded her presentation by encouraging generic pharmaceutical industry to submit alternative BE proposals with modeling and simulation data and to hold early discussions with the Agency via pre‐ANDA meeting and controlled correspondences.

Summary of Presentation “Biopharmaceutics Guidance and Progress on Risk Assessment” by Kimberly Raines, Branch Chief, Division of Biopharmaceutics, Office of New Drug Products (ONDP), Office of Pharmaceutical Quality (OPQ), FDA.

The presentation of Dr. Kimberly Raines focused on the PBPK biopharmaceutics guidance and progress on risk assessment. ⁹ She introduced the concept of PCQS, which is a set of criteria and acceptance ranges to which drug products should conform to deliver the therapeutic benefits indicated in the label allowing to balance of risk/benefit, and patient needs/regulatory challenges. She highlighted PBBM as a promising approach for promoting PCQS through development of predictive in vitro dissolution models based on risk assessment. The capacity of PBBM to incorporate physiological conditions and dissolution information while accounting for relevant physicochemical and physiological factors can lead to a more robust prediction of systemic exposure versus time. As an overview to regulatory utility of mechanistic modeling to support alternative BE approaches, the following objectives were the focus of the presentation ¹ : historical development of the PBPK‐biopharmaceutics application for oral drug product development, manufacturing changes, and controls guidance, including the regulatory impact and current challenges ² ; framework and decision trees portraying initial biopharmaceutics risk assessment in the Knowledge Aided Structured Assessment (KASA) process defining when PBBM approaches have highest regulatory utility (Figure 1 and Figure 2) ⁹ ; and three case studies on how initial biopharmaceutics assessment can demonstrate risk and evaluate how dissolution can provide predictive insights on in vivo performance. ⁹

FIGURE 1.

Biopharmaceutics initial risk assessment for immediate release (IR) solid oral dosage forms (non‐narrow therapeutics index [NTI] or non‐rapid onset drugs). IVIVC, in vitro‐in vivo correlation

FIGURE 2.

Biopharmaceutics initial risk assessment for extended release (ER) solid oral dosage forms (non‐narrow therapeutics index [NTI] drugs). IVIVC, in vitro‐in vivo correlation

In case 1, the example drug product was an immediate release (IR) tablet formulation designed for a BCS class II active pharmaceutical ingredient (API). Based on the biopharmaceutics initial risk assessment decision tree, due to low solubility, high permeability based on absolute bioavailability (BA) of 95% as stated in the reference listed drug (RLD) label, dissolution is not rapid, and identified critical BA attributes (CBAs) that can be adequately detected and controlled, the risk level identified was medium. Therefore, the development of the PBBM at this stage was not required. The data demonstrated that the dissolution method is discriminatory for API particle size, which could directly impact BA of the drug substance. In case 2, the studied drug product was an extended‐release tablet containing a BCS class II API. A dissolution method capable of distinguishing meaningful CBAs was developed. Based on the Biopharmaceutics Initial Risk Assessment decision tree because the in vitro dissolution is independent of the test condition (e.g., medium pH, rotation speed, etc.) and CBAs are clearly identified, the initial risk level identified was medium. Therefore, the development of PBBM at this stage was not required. The proposed dissolution method was discriminative to variations in release‐controlling polymer and additional in vivo data was used to support and determine the appropriate dissolution acceptance criteria based on an established safe space. In case 3, the example drug product was an IR tablet formulation designed for a BCS class IV API. Based on the Biopharmaceutics Initial Risk Assessment decision tree, due to low solubility and low permeability, the originally submitted dissolution method was not sensitive to changes in drug substance or product attributes and did not reject a batch that was not bioequivalent to the phase III clinical batch; therefore, the initial biopharmaceutics risk is high as per the decision tree and PBBM is needed to mitigate the risk. Based on feedback from the Agency related to the initial biopharmaceutics risk the Applicant developed a new in vitro dissolution method and PBBM, through which a clinically relevant in vitro dissolution specification was selected. In summary, the future state of dissolution is an in vitro dissolution test that provides predictive insight to in vivo performance and moving away from dissolution testing solely based on manufacturing capabilities. Predictive in vitro dissolution ensures high‐quality drug products that maintain clinical performance (e.g., safety and efficacy) throughout the product lifecycle. Additionally, the future state of dissolution from a regulatory perspective includes scientific and risk‐based knowledge to support patient‐centric quality standards. Clinically relevant dissolution methods are a crucial element for establishing a link between in vivo performance and the drug product formulation and manufacturing design space and specifications.

Summary of Presentation “Are We Ready to Apply Oral PBPK Modeling for BE Determination?” by Yu‐Chung Tsang, CSO, Biopharmaceutics & Biostatistics from Apotex Inc., Canada.

Dr. Yu‐Chung Tsang talked about “Are we ready to apply oral PBPK modeling for BE determination” from a generic industry perspective. ¹⁰ For market approval of generic drug products, they need to be bioequivalent to the reference product, generally by single‐dose fasted and/or fed PK BE studies. However, BE studies are costly and time‐consuming. Hence, any waiver of in vivo BE studies would provide significant cost and time savings. Currently, the application of oral PBPK modeling for BE determination by the generic industry is rather limited. A common application is for the development of clinically relevant dissolution specifications. An experience of this application for a controlled release (CR) product was shared in the presentation. By applying PBPK modeling using data from in vitro dissolution and in vivo BA studies, dissolution safe space that was shown to provide a PK profile bioequivalent to the reference product was established. With the established dissolution specification limits, bio‐waiver could be granted if future batches of the CR product meet the limits. More applications of oral PBPK modeling for BE determination were also proposed as follows: (1) extend biowaiver to post‐approval manufacturing changes that exceed the Scale‐Up and Post‐Approval Changes level 2 requirements; (2) extend biowaiver to non‐proportionally formulated lower strengths or proportional formulations with dissimilar dissolution profile when BE has been demonstrated with the highest strength; (3) obtain a waiver of fed BE study, if required, after BE has been demonstrated in the fasted state, which is typically considered more discriminatory in detecting differences in BA between formulations or products; and, (4) extend biowaiver to BCS class III drugs that do not have qualitatively (Q1) the same and quantitatively (Q2) similar formulation to the reference product.

The benefit of expanding application of oral PBPK modeling for BE determination is clear; however, there are potential challenges that may need to be overcome, such as lack of confidence in the acceptance of PBPK modeling due to its complexity and uncertainties in the physiological model; subjective simulation findings depending on the modelers; different modeling software may provide different results using the same set of data, as assumptions of some physiological parameters could be different; challenges in conducting PBPK modeling and simulations under fed conditions that can well describe the interaction among food, GI physiology (e.g., GI motility), and drugs; and validation of PBPK modeling could require a large amount of in vivo data, which may compromise some cost and time savings from biowaiver.

Summary of Presentation “Impact of Excipients on Drug Permeation to Support Biowaivers for Non‐Q1/Q2 Products” by Chris Bode, CSO, Vice President, Scientific & Corporate Communications, Absorption Systems.

Dr. Chris Bode introduced their recent research results on the impact of commonly used excipients on permeability of BCS class III drugs, based on a project funded by the FDA contract 75F40119C10127, “Expanding BCS Class III Waivers for Generic Drugs to Non‐Q1/Q2 Products.” ¹¹ Currently, the FDA and the European Medicines Agency (EMA) allow biowaiver of PK BE studies for BCS class I drugs (highly soluble and highly permeable) and for some BCS class III drugs (highly soluble and low permeable). However, the current criteria for excipient similarity for BCS class III drug products are quite stringent, which has led to underutilization of the BCS biowaiver pathway for this class of drugs. In addition, there is a need for a robust, predictive in vitro model for evaluating the effects of excipients on drug permeation. To address both of these issues, the in vitro dissolution absorption system (IDAS) was used to evaluate the effects of common excipients on the permeation of a cassette of five model drugs across Caco‐2 cell monolayers. ¹¹ The IDAS is comprised of a dissolution vessel containing two permeation chambers, with polarized monolayers of Caco‐2 cells mounted in the interface between the two. Dissolution of a drug product is monitored by measuring the amount of dissolved API in the dissolution vessel (donor chamber) over time, and API permeation across Caco‐2 cell monolayers is assessed by measuring its rate of appearance in the permeation (receiver) chamber. Four of the model drugs (acyclovir, atenolol, cimetidine, and ranitidine) belong to BCS class III, and one model drug (minoxidil) belongs to BCS class I. With IDAS, it is possible to measure drug dissolution and permeation simultaneously from solid oral dose forms. However, in this phase of the study, the model drugs were pre‐dissolved; therefore, only permeation was measured.

Fifteen excipients, representing multiple functional classes, were evaluated one at a time across a range of pharmaceutically relevant concentrations. Overall, there were few examples of excipients producing obvious and dose‐dependent effects on permeation of any of the model drugs, and even fewer with effects on all of the class 3 model drugs. Seven of the excipients tested had no effect on the permeation of any of the model drugs: hydroxypropyl methylcellulose (at two viscosities), microcrystalline cellulose, croscarmellose sodium, talc, mannitol, and silicon dioxide. Another seven excipients had some effects, generally not dose‐dependent, on only one or two of the model drugs. Sodium lauryl sulfate, a known permeation enhancer in vitro, ¹² was the only excipient that caused dose‐dependent increases in the permeation of all model drugs (BCS class I and III). Although Caco‐2 cells are known to be overly sensitive to excipients compared to intestinal epithelium in vivo, this appears not the case in the IDAS, perhaps because the Caco‐2 cell monolayers are oriented vertically, rather than horizontally as in the more conventional transwell format. ¹² , ¹³ The results suggest that, with regard to BCS‐based biowaivers for class 3 drugs, it may be possible to relax the current criteria for excipient similarity for many excipients without affecting in vivo BE. This could have important consequences for the development and regulatory approval of generic drugs.

PANEL DISCUSSION FOR DAY 2 SESSION 1

This session was moderated by Hongling Zhang, acting division director from the Office of Bioequivalence, OGD, Center for Drug Evaluation and Research (CDER), and Tycho Heimbach, Senior Principal Scientist/Director from Merck & Co., Inc. The panelists included Fang Wu from OGD, FDA; Kimberly Raines from ONDP, FDA; Chris Bode from Absorption Systems; Yu Chung Tsang from Apotex; Amitava Mitra from Janssen R&D; James Polli from University of Maryland; Liang Zhao from OGD, FDA; and Yi Zhang from OGD, FDA.

In the panel discussion, the following questions were discussed and the comments from panelists are summarized below:

Question 1

What are common reasons for non‐acceptance of PBPK based biowaivers?

The panel provided the following common reasons for non‐acceptance of PBPK based biowaivers: (1) model is not sufficiently validated for its intended purpose/context of use; (2) the model inputs, which include dissolution or solubility, are not biorelevant/biopredictive; (3) lack of justification for input model parameters; (4) for the models used to support biowaiver, lack of non‐BE or unacceptable BA batch to challenge the model and increase confidence in the model and the tested range/space of the formulation variations; (5) insufficient in vivo PK datasets for construction, verification or qualification of the model (e.g., lack of independent in vivo PK data, or lack of human i.v. data to estimate the drug disposition, or lack of oral solution data with complete absorption as it is recognized that human i.v. data generation may not always be feasible); and (6) formulation variations included in model verification are not wide enough leading to extrapolation outside of the tested space of the formulation variations, particularly for the purpose of a biowaiver or justification for specifications.

Question 2

If PBBM/PBPK modeling is used to justify the ranges of drug product specifications via virtual BE, what should be used as the reference, the profile for upper bound, the target profile for the test product, or the target profile for the RLD, etc.?

The panel pointed out that reference selection is case‐specific. If it is for post‐approval change, the target profile for the approved product may be used. If it is for bridging or some other issues, profile of reference product may be used. The safe space of critical quality attributes can be supported by virtual BE (VBE) trials. ² It would be recommended to leverage data from different formulations (e.g., varying release rates). Therefore, early communication with the Agency is recommended. Additionally, for justification of specifications, it would be important to have enough formulation variance being tested and appropriately put into the model. VBE testing should be conducted to identify a range of virtual dissolution profiles that are bioequivalent to the pivotal clinical batch.

Question 3

The IDAS dissolution permeability system was used to evaluate the impact of excipients on permeability. Are these in vitro results translatable to in vivo performance? Can the sponsor use IDAS to demonstrate that a given excipient does not impact permeability to support a study waiver?

According to the panelist, the IDAS with Caco‐2 cells is less sensitive to excipients than those in the standard transwell system. IDAS is a static in vitro system and exposes the drug and excipients to the Caco‐2 cells long enough (1 to 2 h, as opposed to a transient exposure of a given cell in vivo) to test the worst‐case scenario. The extended exposure time in vitro is necessary to allow measurable amounts of drug to permeate to the receiver side of the cell monolayer. Therefore, if IDAS demonstrates no effect of excipients on the permeability of a BCS class III drug even under such worst‐case conditions, the results may be translatable.

Question 4

Please comment on the use of biorelevant dissolution data to support PBPK modeling and how is biopredictivity established for dissolution and permeability data

The panelist commented that the dissolution profiles in the case example mentioned in Dr. Tsang's presentation ¹⁰ were generated using USP method for quality control purpose. The involved drug is highly soluble. As the drug was formulated as a CR product, the in vivo drug release could still be shown to be dissolution‐dependent. This was seen via the rank correlation of the in vitro release rate and maximum concentration in the BE study. Hence, although biorelevant media were not used for generating dissolution profiles, biopredictability was still established. This indicates that the use of biorelevant dissolution medium may be preferred but may not be a necessity for some cases.

Question 5

What are the considerations for validating a PBPK absorption model to support biowaivers?

The panel proposed that when considering the validation ² , ¹⁴ of a PBPK absorption model, the predictive performance of a PBPK model should be validated for its intended purpose. Independent datasets that were not used in model development are recommended to challenge the predictive performance of the model. To increase confidence in the model, it is recommended that applicants use PK data from batches exhibiting non‐BE or unacceptable BA to challenge the model and demonstrate the model's predictive performance and increase confidence on the established PBPK model. Using datasets obtained from different doses and different release rate are recommended to establish the in vitro and in vivo relationship.

Question 6

What can be done to further promote and encourage the use of PBPK by innovators/generics?

Per the panel discussion, the Agency is promoting and supporting modeling approaches, which should be fit for purpose and risk‐based. For example, sensitivity analysis can be conducted to predict the worst‐case scenario of the excipient effects on permeability and to evaluate whether this would pose a risk of non‐BE. Industry stakeholders are encouraged to use creative and implementable modeling approaches. The following areas of PBPK modeling, including biowaiver for post‐approval changes, biowaiver for non‐compositionally proportional formulations, waiver for fed BE studies, and biowaiver for BCS class III drugs would be the impactful topics. Further, the panel also pointed out that building a database with model input parameters (e.g., physicochemical properties, absorption and disposition parameters of drugs, and effects of excipients) would be helpful for minimizing the differences of modeling and simulation outcomes for the same drug caused by different modelers and software. More education and communication among the pharmaceutical industry, academia, and regulators are needed regarding how PBPK models are developed and validated and why some assumptions may not affect the modeling outcome.

PRESENTATIONS FOR DAY 2 SESSION 3

Speakers from session 3 provided discussions and case examples on integrating biopharmaceutic data and GI physiology in PBPK models and on developing PBPK models (adult and pediatric populations) to determine dissolution safe space. Over time, the use of modeling approaches has allowed the true complexity of in vitro experiments to be appreciated and more fundamental parameters to be derived. The mechanistic analysis of in vitro biopharmaceutics experiments helps in confirming and/or estimating intrinsic parameters required for mechanistic simulations of in vivo performance of orally administered drug products. With population PBPK modeling platforms, between‐subject variability in physiological parameters is also taken into consideration, allowing the estimation of the variability around drug‐related parameters required for simulations in vivo. However, navigating the multiplicity of in vitro oral biopharmaceutics models might be challenging, and a best practice framework is yet to be developed. On the other hand, mechanistic absorption/PBPK models have recently been used to support formulation development in areas such as drug discovery, clinically relevant drug product specifications (e.g., dissolution and particle size specification settings), quality risk assessment and in vivo BE simulations in adults. ⁷ Based on the successful prediction of drug disposition in the GI tract and plasma drug concentration‐time profiles in adults, development of pediatric PBPK models appear to be feasible. However, there are several challenges related to developing PBPK models and their application in pediatrics.

Summary of Presentation “Integrating Biopharmaceutic Data and Gastrointestinal Physiology Using Mechanistic Modeling” by Rodrigo Cristofoletti, Assistant Professor, Center for Pharmacometrics & Systems Pharmacology, Department of Pharmaceutics Lake Nona (Orlando), University of Florida.

Dr. Rodrigo Cristofoletti presented on the integration of biopharmaceutics data and GI physiology using PBPK modeling. ¹⁵ The use of modeling approaches has explored true complexity of in vitro experiments and derived fundamental in vitro parameters. Such model‐based analysis of in vitro data allows one to separate drug‐specific and in vitro system parameters (i.e., deconvolution of in vitro data). Similar concepts can be applied to in vitro biopharmaceutics experiments in the form of mechanistic analysis of such assays to confirm and/or estimate intrinsic parameters required for mechanistic oral simulations of in vivo performance taking in consideration between‐subject variability in physiological parameters. Two PBPK modeling case studies were presented. The first focused on applying a reverse translation approach to identify a biopredictive dissolution method for ibuprofen (a BCS class II weak acid). A point‐to‐point, mechanistic in vitro‐in vivo correlation was established for ibuprofen after using a phosphate buffer concentration able to emulate the microclimate pH surrounding ibuprofen particles dissolving in physiological bicarbonate buffer. ¹⁶ , ¹⁷ The model successfully predicted the anticipated test/reference geometric mean ratios of the BE metrics; however, the respective 90% confidence intervals were rather narrow due to the knowledge gap related to predicting within‐subject variability in physiological parameters within PBPK modeling platforms. The second example illustrated the application of PBPK modeling to predict the PK profile of ketoconazole (BCS class II weak base) using in vitro precipitation kinetics from dumping test, transmembrane flux, and biphasic dissolution. The model successfully predicted fractional precipitation in duodenum and the PK profile of ketoconazole oral solution using precipitation kinetics derived from biphasic dissolution experiments. However, the model failed to reproduce ketoconazole systemic exposure at higher doses, which suggests more research on using mechanistic models based on classical nucleation theory is needed.

Summary of Presentation “Modeling for Success: A Case Example for Oseltamivir Phosphate” by Youssef Mousa, Pharmacologist from ORS, OGD, FDA.

Dr. Youssef Mousa talked on the challenges in developing a pediatric PBPK model from adult model and the use of PBPK modeling to support formulation development in areas such as drug product quality perspective (e.g., dissolution specifications). ¹⁸ Generally, the pediatric PBPK model follows adult model development and evaluation. This approach requires complete understanding of drug absorption, distribution, metabolism, and excretion (ADME) in adults before model extrapolation to pediatrics. ¹⁹ However, there are several challenges related to developing PBPK models and their application in pediatrics. These challenges include the following factors and assumptions: the pathways of clearance are similar in adults and children, ontogeny factors are based on healthy subjects, variability in terms of physiological factors is similar between adults and pediatric patients, limitations of allometric scaling, drug‐specific parameters (e.g., solubility and dissolution) remain constant regardless of the difference in adult versus pediatric GI physiology, and the ontogeny of different physiological factors. The presented example focused on using PBPK modeling to determine the dissolution boundary to maintain BE between the test and reference oseltamivir phosphate (OP) drug products using PBPK modeling‐based VBE simulations in adults and pediatrics. The adult and pediatric (at different age groups) models were validated with rich in vivo PK data and were used to determine the dissolution boundary to maintain BE between the test and reference OP products in adults and 0–2 months and 9–19 years pediatric populations by performing VBE simulations. This example shows the applicability of PBPK modeling in mitigating risk of formulation/batch changes and providing a quantitative basis for setting clinically relevant dissolution specifications in both adults and pediatrics.

PANEL DISCUSSION FOR DAY 2 SESSION 3

This session was moderated by Lanyan (Lucy) Fang, deputy division director from ORS, OGD, FDA, and Filippos Kesisoglou from Merck & Co., Inc. The panelists include Rodrigo Cristofoletti from University of Florida; Youssef Mousa from ORS, OGD, FDA; Jianghong Fan from Office of Clinical Pharmacology, FDA; Tycho Heimbach from Merck & Co., Inc.; Myong‐Jin Kim from OGD, FDA; Duxin Sun from University of Michigan; Yu Chung Tsang from Apotex; and Banu Zolnik from Office of Pharmaceutical Quality, FDA.

In the panel discussion, the following questions were discussed and the comments from panelists are summarized below:

Question 1

In the case example of OP, PBPK modeling and simulation was used to determine bioequivalent dissolution “Safe Space” for both adults and pediatrics. The modeling includes interpolation or extrapolation of model behavior outside formulations for which there are dissolution and clinical data. Please comment on the considerations when we extend this framework/strategy to other products with confidence?

PBPK modeling is a much‐needed area for testing BE to support generic drug approval in the pediatric population. PBPK modeling has been extensively used in the new drug area and it supported the approval process for generic drugs and with pediatric population. This kind of modeling has improved the review process over the past decade in pharmacometrics, pharmacogenomics, and clinical pharmacology in the new drug area. In addition, PBPK modeling can be used to support BE evaluation in pediatrics under the age of 6 years, especially if there is risk in extrapolating the BE results from adults to pediatrics. Generally, if a difference in absorption between generic and brand drug products is not observed in adults, it is unlikely that such a difference will be observed in pediatrics. For developing a pediatric PBPK model, it is recommended to use bio‐predictive dissolution data generated in bio‐relevant media and to consider the ontogeny of metabolizing enzymes and transporters and many other factors (e.g., food effect, gastric pH, and fluid dynamics in the GI tract) that may differ between adults and children. This approach would increase the confidence in the developed model, with sufficient verification and validation. The example of developing a PBPK model for OP is the way to look at formulation performance and BE modeling from adults all the way to neonates. However, for other drug products, an important factor to consider is the availability of rich in vivo PK data to validate the pediatric model at different age groups. Such an approach can only be applied to an age group with sufficient size range combined with sufficient information about ontogeny of this population. In addition, caution should be considered when pediatrics get lumped together as pediatrics have different age groups with different ADME characteristics that may need to be included in PBPK model building. For the case example of OP, the exploration of generating additional hypothetical dissolution profiles by shifting the release profile rather than decreasing the percentage, and to use different virtual sample size were recommended.

Question 2

For translational research, how does one select an in vitro system to predict absorption or precipitation kinetics for modified release or immediate release formulations with BCS class II or IV compounds and what is the best way to incorporate them in the model?

From modeling exercises, different in vitro setups may give different results. Closed in vitro systems, like the transfer model or dumping test, tend to overestimate drug precipitation kinetics. Therefore, more advanced in vitro systems like the biphasic or overall systems that consider drug removal from solution by absorption/permeation may show better performance to predict weak base drug behavior in the intestinal lumen. However, it is challenging to generalize these findings for other weak base drugs such as itraconazole. Therefore, conducting reverse translation exercises would breach some knowledge gaps in the translatability of in vitro data, especially in vitro precipitation data.

Precipitation rate kinetics can be better estimated by accounting for drug‐dependent permeation and precipitation behavior. However, at a higher level, the in vitro and in vivo release of modified release or immediate release formulations with BCS class II or IV compounds are important factors in predicting the absorption and precipitation kinetics. There are several challenges associated with the in vitro dissolution conditions, such as pH buffer capacity, media volume, and rotation speed. Gastric transit time is considered an important physiological factor in dictating dissolution time, especially for weak acid drugs, and predicting the PK profile. From a PBPK modeling perspective, many factors, such as precipitation, volume of each GI compartment, transit time, buffer capacity, and surfactant in the small intestine, should be considered when modeling weak base drugs. However, for weak acid drugs, stomach factors such as gastric transit time should be considered specifically.

Question 3

For pediatric products, PBPK approaches have been suggested to bridge PKs of different formulations in pediatrics as it can integrate the interplay between physiology and formulation factors. Please comment on the key factors that the PBPK models should incorporate for this intended purpose.

To develop a pediatric PBPK model with confidence, it is necessary to identify the potential model limitations and the uncertainties when extrapolating the adult model to pediatric model. It is important to consider the following: first, drug‐dependent properties such as physicochemical properties, log P, pKa, solubility (i.e., change in BCS class), precipitation or supersaturation, stability in the GI tract, susceptibility to metabolism enzymes and transporters in the GI tract, special formulation excipients that may affect the absorption process, and whether there is a special release mechanism; second, the type of individual dissolution approaches used in model building; and, third, whether the age‐dependent parameters have been already incorporated in the model or not with sufficient verification. The bulk of this information may help in identifying the rate‐limiting step(s) in the drug absorption process; therefore, recognizing which kind of parameter is connected to the rate‐limiting step. If those parameters are not related to the drug product quality, then there is a certain confidence level in regard to the model performance; otherwise, more data are needed to validate the model. The validation of a pediatric model at different age groups (e.g., 6–18 years, 6–12 years, or 12–18 years) is limited by the scarce in vivo PK data in some cases. This would introduce challenges in assessing the effect of formulation changes on BE in pediatric population. However, with the help of PBPK modeling along with the efficacy and safety data, the effect of formulation changes and risk of non‐BE may be evaluated.

COMMON TOPICS COVERED IN ORAL PBPK SUBMISSIONS AND REASONS FOR FAILING

It is worth noting that during meeting registration, the workshop organizer conducted a survey on the following questions related to the above topics: What were the topics covered by your oral PBPK submissions in NDAs? What were the topics covered by your oral PBPK submissions in ANDAs? For modeling submissions which were not successful, what was the main reason?

The survey results are shown in Figures 3, 4, 5. Based on the responses from the registered people in this workshop, the survey revealed that the most common topics covered in oral PBPK submissions in NDAs are waiver of in vivo PK BE study, and justification of drug substance particle size specifications and dissolution safe space. In ANDA submissions, these are also the common topics. The top five reasons for modeling submissions that were not successful include: model was not sufficiently validated, model was not fit for purpose, model was predicting no variation or knowledge space was not large enough, model was not parameterized adequately, and clinical variants tested were not representative of the question.

FIGURE 3.

Survey results for what were the topics covered by your oral PBPK submissions in NDAs. NDA, new drug application; PBPK, physiologically‐based pharmacokinetic; PK, pharmacokinetic

FIGURE 4.

Survey results for what were the topics covered by your oral PBPK submissions in ANDAs. ANDAs, abbreviated new drug applications; PBPK, physiologically‐based pharmacokinetic; PK, pharmacokinetic

FIGURE 5.

Survey results for what was the main reason for modeling submissions which were not successful

CONCLUSIONS

Use of modeling and simulation approaches, including PBPK, as an alternative BE approach and as a tool for risk assessment and biowaiver are increasingly seen in regulatory submissions. This workshop built a venue for the community to establish collaboration and share successful cases and views on the challenges and opportunities on the utility of oral PBPK in regulatory submission.

FUNDING INFORMATION

No funding was received for this work.

CONFLICT OF INTEREST

The authors declared no competing interests for this work.

Wu F, Mousa Y, Raines K, et al. Regulatory utility of physiologically‐based pharmacokinetic modeling to support alternative bioequivalence approaches and risk assessment: A workshop summary report. CPT Pharmacometrics Syst Pharmacol. 2023;12:585‐597. doi: 10.1002/psp4.12907