Physiologically Based Pharmacokinetic Modeling and Simulation in Regulatory Review: US FDA CBER Experience and Perspectives

Abstract

Physiologically based pharmacokinetic (PBPK) modeling has emerged as a valuable tool in model-informed drug development (MIDD). This approach enables the integration of diverse experimental data to predict pharmacokinetics (PK) and dosing regimens and facilitates understanding of mechanism of action (MoA) and pharmacodynamics (PD). In this article we provide a landscape analysis of PBPK submissions at the U.S. Food and Drug Administration, Center for Biologics Evaluation and Research (CBER). We summarize CBER’s experience on PBPK modeling and simulation (M&S) for therapeutic proteins, cell and gene therapy products. We discuss specific case studies that illustrate the use of PBPK for dose selection of therapeutic proteins, highlight recent progress and provide our perspectives on potential application of PBPK for adeno-associated virus (AAV)-based gene therapies and messenger RNA (mRNA) therapeutics. For cell and gene therapy products, PBPK M&S is emerging as MIDD approaches to support clinical trial design, dose selection, predicting PK/PD, and facilitate quantitative understanding of safety and efficacy. As the field continues to evolve, PBPK modeling is well positioned to provide supportive evidence to facilitate the development of safe and effective biological products.

Introduction

Model-informed drug development (MIDD) involves developing and applying exposure-based, biological and statistical models to facilitate drug development and regulatory decision-making for small molecule drugs, therapeutic proteins, cell and gene therapies (CGTs).^[1] Among such MIDD approaches gaining significant attention in the regulatory settings is physiologically based pharmacokinetic (PBPK) modeling and simulation (M&S). PBPK modeling is a mathematical framework that describes the absorption, distribution, metabolism, and excretion (ADME) of a drug based on its physicochemical and biochemical properties, along with physiological parameters, including disease-related parameters. Application of PBPK models in drug development allows the integration of various data types (e.g., in vitro, preclinical animal, physiological, clinical trial, and real-world data), which enables prediction of pharmacokinetics (PK) and facilitates understanding of mechanism of action (MoA), pharmacodynamics (PD), efficacy, and safety.

The past decade has seen significant advances in basic sciences and regulatory initiatives that facilitate the application of MIDD, including PBPK modeling, in drug development and regulatory decision-making.^[2–6] The development and application of PBPK models for biological products (e.g., therapeutic proteins, vaccines, CGTs) is challenging in part due to complex product related attributes, immunogenicity, disease-context and patient-specific factors. Scientific advances in molecular biology, bioinformatics, pharmacology, and data analytics have contributed to our understanding of diseases pathophysiology and MoA, identification of biomarkers, and development of predictive quantitative models. From the regulatory perspective, the FDA MIDD meeting program that started as pilot and later transitioned to MIDD paired meeting program has created engagement opportunity to discuss the application of MIDD approaches in drug development and regulatory evaluation.^[7] The FDA published a white paper on the use of risk-based credibility assessment for PBPK^[8] and conducted a dedicated workshop on development of best practices in PBPK to support clinical pharmacology regulatory decision-making.^[9] The risk-based credibility framework is now being incorporated for wider application of MIDD evidence generation as part of the ICH guideline on “General Principles for Model-Informed Drug Development – M15”.^[10] In April 2025, FDA released a roadmap that outlines opportunities to reduce animal testing during drug development. In particular, new approach methodologies (NAMs) were described, and specifically mention PBPK, quantitative system pharmacology (QSP) and machine learning/artificial intelligence predictive models to leverage existing data to predict safety, immunogenicity and pharmacokinetics reducing the need for new animal experiments.^[11] This provides yet another opportunity for PBPK models to support regulatory decisions, which can inherently provide drug distribution and toxicity predictions as part of its modeling framework.

In this article we summarize Center for Biologics Evaluation and Research (CBER) experience, discuss specific case studies and provide perspectives on PBPK modeling for therapeutic proteins, adeno-associated virus (AAV)-based gene therapies (GTs), and messenger RNA (mRNA) therapeutics.

Overview of CBER Experience on PBPK modeling and Simulation

We conducted a comprehensive landscape review of electronically submitted regulatory applications to FDA’s CBER to examine the current state of PBPK M&S. PBPK modeling played a role in 26 regulatory submissions and interactions from 2018 and 2024 (Figure 1), supporting applications from 17 sponsors for 18 products, 11 of which are intended to treat patients with rare diseases. This selected timeline aligns with FDA’s effort in advancing MIDD as part of Prescription Drug User Fee Act VI and VII. The number of interactions increased over time (Figure 1A). The submissions included 1 Biologics License Application (BLA), 10 Investigational New Drug (IND) applications, 8 pre-Investigational New Drug (pre-IND) meetings, and 7 others, including meetings for Initial Targeted Engagement for Regulatory Advice on CBER/CDER Products (INTERACT), MIDD program, and Type V Drug Master File submissions (Figure 1B). The products associated with these submissions include gene therapy products (8), plasma-derived products (3), vaccines (1), cell therapy product (1), and other products (5 submissions that includes small molecules and bacterial lysates) (Figure 1C). The majority of the proposed PBPK models are designed to justify and optimize drug dosing in early development, offer a mechanistic understanding of how drugs are processed and interact with targets in the body, and guide dosing strategies for specific populations (e.g. pediatrics). Additionally, some models specifically address the prediction and evaluation of Drug-Drug Interactions (DDIs).

Figure 1:

Analysis of CBER-Industry regulatory interactions involving PBPK models (A) annual number of industry-CBER regulatory interactions on PBPK, (B) showing counts for IND, pre-IND, INTERACT, MIDD, BLA, and Type V Drug Master File on PBPK, and (C) number of PBPK submission for product categories including Gene therapy products, Plasma-derived products, Vaccines, Cell therapy product and Others.

In subsequent sections, we describe specific case studies for therapeutic proteins and discuss opportunities and challenges for applying PBPK modeling for AAV-based GTs and mRNA therapeutics.

Therapeutic Proteins

Plasma-derived and related recombinant products are therapeutic proteins primarily approved to replace physiologically low levels or absent proteins in humans. For example, treatment of hemophilia A (congenital Factor VIII deficiency) focuses on prophylactic intravenous (IV) infusion of either plasma-derived or standard or extended half-life recombinant Factor VIII (rFVIII) products. A major limitation of these Factor VIII therapies is the short half-life, which requires 3-4 times weekly infusion (standard half-life) and 2-3 times weekly infusions (extended half-life). FVIII is a large heterodimeric protein circulating in plasma in complex with von Willebrand factor (VWF). The fraction of FVIII bound to VWF is cleared via VWF clearance mechanisms with the plasma half-life of ~20 hours, which results in a biochemical ceiling effect for traditional approach of half-life extension of FVIII.^[12] The FDA has recently approved ALTUVIIIO [recombinant antihemophilic factor Fc-VWF-XTEN fusion protein-ehtl] for use in adults and children with hemophilia A. ALTUVIIIO has 3 to 4-fold prolonged half-life relative to other standard and extended half-life FVIII products, which enable once weekly infusion.

ALTUVIIIO is a recombinant FVIII analogue fusion protein that is independent of endogenous VWF to overcome the half-life limit imposed by the interaction of FVIII with VWF.^{[12, 13]} Appending the D’D3 domain of VWF to a recombinant FVIII-Fc fusion protein provides protection and stability to FVIII, and prevents FVIII interaction with endogenous VWF, thus overcoming the limitation on FVIII half-life imposed by VWF clearance.^{[12, 13]} The Fc region of human immunoglobulin G1 (IgG1) binds to the neonatal Fc receptor (FcRn). FcRn is part of a naturally occurring pathway that delays lysosomal degradation of immunoglobulins by recycling them back into circulation, thus extending the plasma half-life of the fusion protein.^{[12, 13]} ALTUVIIIO contains two XTEN polypeptides, which alter the hydrodynamic radius of the fusion protein, thus reducing rates of clearance and degradation, and improving PK properties.^[14] Given the expected changes to clearance and distribution of the FVIII replacement therapy, using multiple MIDD approaches that incorporate prior knowledge and experience from similar products would have assisted in the regulatory decision-making for this product.

MIDD approaches such as population pharmacokinetic (popPK) and exposure-response analysis are often performed in development and regulatory review of plasma-derived and related recombinant products.^[15] PopPK and exposure-response analysis were also included in regulatory assessment of ALTUVIIIO.^[16] Furthermore, the regulatory dossier for ALTUVIIIO includes a PBPK model which was primarily developed to support dose selection for pediatric patients <12 years of age. Based on modeling and simulation, maintaining FVIII activity >40 IU/dL for a dosing interval is expected to deliver an optimal effect in reducing the bleeding risk; meanwhile, maintaining FVIII activity >20 IU/dL for a dosing interval is deemed sufficient. In pediatrics younger than 12 years of age, while FVIII activity is only maintained above 40 IU/dL for 35-43% of a dosing interval, the effect in bleeding prevention could still be adequate since the FVIII activity >20 IU/dL is maintained for majority of a dosing interval.

A minimal PBPK model structure for monoclonal antibody (mAb) was employed to describe the distribution and clearance mechanism involving FcRn recycling pathway. The PBPK model was first developed and evaluated in adult and children using clinical data obtained from FDA approved product, ELOCTATE (Fc Fusion Protein, rFVIII product). ELOCTATE data was leveraged to validate the pediatric PK prediction since it also contains an Fc fusion protein and its clearance mechanism involves FcRn in part. For pediatric patients the effects of age on FcRn abundance and vascular reflection coefficient were optimized using clinical PK data from ELOCTATE. The PBPK model for ELOCTATE predicted the maximum concentration (C_max) and area under the curve (AUC) values in both adult and children with reasonable accuracy (prediction error within ±25%; Table 1) suggesting that the model was able to describe the FcRn mediated recycling pathway.^[16]

Table 1:

Observed and PBPK Predicted Mean PK Parameters of Eloctate and ALTUVIIIO in Adults and Pediatrics after Single IV Dose Administration.

Population	Age (years)	Drug	Dose (IU/kg)	Cmax (ng/mL)			AUC (ng.h/mL)
				Observed	Predicted	%Error	Observed	Predicted	%Error
Adult	23-61	ELOCTATE	25	140	105	−25	3,009	2,671	−11
			65	345	272	−21	7,794	6,944	−11
	19-63	ALTUVIIIO	25	282	288	2	14,950	13,726	−8
			65	735	749	2	43,300	35,687	−18
Pediatric	≥2 to <6	ELOCTATE	50	174	196	12	2,422	2,915	−2
	≥6 to <12			193	188	20	3,244	3,029	−7

Note: AUC values for ALTUVIIIO are AUC_0-360h and for Eloctate AUC are AUC_0-168h. Data sources ^[16].

The PBPK model for ALTUVIIIO predicted the C_max and AUC values with <20% prediction error in adult population (Table 1). Pediatric PBPK simulations were conducted with 10 virtual trials of 10 participants/age group in each virtual simulation (n = 100) following ALTUVIIIO IV once weekly (QW) doses of 50, 65 or 80 IU/kg for 4 weeks. The PBPK simulation showed that in children (<12 years of age), the ALTUVIIIO dosing regimen of 50 IU/kg IV once weekly (QW) maintain FVIII activity levels above 10 IU/dL for 2-5 days and achieve mean trough FVIII activity levels between 2.6-5.6 IU/dL.^[16]

Considering that there is limited knowledge on quantitative information regarding the ontogeny, FcRn recycling pathway for Fc fusion proteins and clearance mechanism for FVIII-VWF, the PBPK analysis were used as supplementary support for pediatric dose selection for the Phase 3 study.^[16] Overall, although plasma-derived and related recombinant proteins such as blood clotting factors have been regularly submitted to CBER, this is the first case that uses PBPK M&S as supporting information for dose selection in a BLA submission. CBER review staffs are also evaluating and engaging in sponsor meetings discussing PBPK M&S for other plasma-derived products including polyclonal antibodies (Figure 1).

AAV-based Gene Therapies

AAV-based GTs are viral vectors that have been genetically engineered to be replication deficient and express a transgene of interest in vivo. AAVs are traditionally classified into one of nine different serotypes based on the composition of the proteins on the capsid, which also contributes to the AAV tissue tropism.^[17] Genetic engineering of the capsids has also expanded the classification beyond the traditional serotypes, with the primary aim of reducing the immunogenicity, enhanced transduction and specific tissue tropism of the vector. There are currently 8 approved AAV GT products (Table 2).^[18] Each product was assessed for the vector biodistribution and shedding in all detectable tissues, as well as pharmacodynamic endpoints, when applicable, and relationships between patient characteristics and PK/PD were explored during the review process (Table 2).

Table 2.

Summary of FDA Approved AAV Gene Therapies.

Product Name (Initial Approval Year) [Citation]	Route of Administration	Vector Type	Dose(s)	PK Endpoint (Assessed tissue biodistribution and shedding of vector DNA)	PD Endpoint
Beqvez (fidanacogene elaparvovec-dzkt) (2024) [52]	IV	AAV serotype Rh74	Single dose 5x1011 vector genomes/kg	Plasma, saliva, urine, peripheral blood mononuclear cells, and semen	Serum factor IX (FIX) activity
Kebilidi (eladocagene exuparvovec-tneq)* (2024) [53]	Intraputaminal infusion	AAV serotype 2	4 total infusions totaling 1.8 x1011 vector genomes	Putamen, cerebellum, cerebrum, spinal cord, CSF, serum, urine	Production of homovanillic acid in cerebrospinal fluid
Elevidys (delandistrogene moxeparvovec-rokl)* (2023) [54]	IV	AAV serotype rh74	Single dose <70kg: 1.33x1014 vector genomes/kg; >70kg: 9.31x1015 vector genomes/kg	Muscle, serum, saliva, urine and feces	Micro-dystrophin protein expression from muscle biopsies
Roctavian (valoctocogene roxaparvovec-rvox) (2023) [55]	IV	AAV serotype 5	Single dose 6 x 1013 vector genomes/kg	Serum, semen, liver, lung, heart, lymph nodes, kidney, spleen, bone marrow, testis, brain	Serum factor VIII (FVIII) activity
Adstiladrin (nadofaragene firadenovec-veng) (2022) [56]	Intravesical Instillation	Adenoviral serotype 5	3x1011 viral particles/mL, once every 3 months	Blood, urine	Human interferon alpha-2b (IFNα2b) concentration in urine
Hemgenix (etranacogene dezaparvovec-drlb) (2022) [23]	IV	AAV serotype 5	Single dose 2 x1013 genomic copies/kg	Serum, liver, adrenal glands, saliva, nasal secretions, semen, urine, feces	Serum FIX activity
Zolgensma (onasemnogene abeparvovec-xioi) (2019) [57]	IV	AAV serotype 9	Single dose 1.1 x 1014 vector genomes/kg	Saliva, urine, stool, liver, spleen, heart, pancreas, lymph node, skeletal muscles, peripheral nerves, kidney, lung, intestines, gonads, spinal cord, brain, thymus	Immunostaining for SMN protein in deceased patients showed generalized SMN expression in spinal motor neurons, neuronal and glial cells of the brain, and in the heart, liver, skeletal muscles.
Luxturna (voretigene neparvovec-rzyl) (2017) [58]	Subretinal injection	AAV serotype 2	Single injection for each eye, 1.5 x 1011 vector genomes	Intraocular fluids, optic nerve, optic chiasm, spleen, liver, lymph nodes, tears, serum	Retinoid isomerohydrolase RPE65 expression in subretinal space

^*=Denotes Accelerated Initial Approval.

IV = Intravenous infusion, AAV = Adeno-associated virus.

Efforts to advance the application of MIDD into AAV GTs have focused on understanding the biological processes in model organisms that allows for extensive sampling from various tissue compartments and link the disposition of the AAV and the transgene (e.g. monoclonal antibody) via a transgene production process.^[19] Parsimonious approaches have also been proposed, such as exploring an allometric link between the gene expression efficiency observed in model organisms to guide human dose selection.^[20] However, clearly established trends across therapeutic contexts remain elusive, with dosing targets ranging many orders of magnitude, and different considerations are often involved in the delicate benefit-risk trade-offs,^[21] such as being under-dosed for therapies that can only be administered a single time, or the inflammatory responses and liver toxicities associated with doses that may be too high.

PBPK models can incorporate mechanistic concepts from different target patient populations or disease contexts. For instance, when evaluating the potential effects of age, sex and comorbidities on PK/PD of AAV GTs, it may be informative to perform a sensitivity analysis to explore the impacts of varying physiological factors, such as higher hepatic cell turnover and dilution rate of AAV in growing organs in pediatric patients.^[17] This type of mechanistic information can also be incorporated into PBPK models to estimate and project the durability of the PD activity. For instance, during the review of HEMGENIX, an AAV- based gene therapy designed to deliver a copy of a gene encoding the Padua variant of human coagulation Factor IX (FIX) to increase circulating FIX activity and prevent bleeding in adult patients with Hemophilia B, we noted a trend for impact of age on FIX activity. Although there were no pediatric subjects, a trend was observed with age, in that those <40 years old had 1.5-2-fold lower FIX activity levels when compared to those >60 years.^[22] In addition, individuals with elevated alanine aminotransferase (ALT) levels (13/53) had approximately 44% lower mean FIX activity 18 months after treatment compared to those that did not have elevated ALT. The effect of corticosteroids on FIX durability were also explored and again illustrates the complex biochemical and physiological considerations that need to be considered when analyzing time-series data for patients treated with AAV GTs.^[23]

With sufficient data, PBPK models could be employed for many AAV-based GTs to predict biodistribution, PK and transgene concentrations. In this respect a validated PBPK model has the potential to reduce patient burden or replace some aspects of the PK and biodistribution analysis in rare disease patient populations. For example, in review of HEMGENIX we noted that some patients were unable to provide blood and semen (to monitor the shedding of the viral vector in patients with hemophilia B) samples during multiple clinical visits, and it was challenging to determine the clinical definition of “complete clearance” of viral vector per protocol. In the clinical efficacy study (N = 54), a total of 56% (30/54) of subjects achieved absence of vector DNA from blood and 69% (37/54) from semen by month 24. Several subjects did not return the required number of blood and semen samples to assess the shedding status as per the definition of three subsequent measurements below limit of detection of vector DNA.^[23] PBPK M&S also have the potential to predict DDIs. For example, clinical studies and in vitro studies demonstrated potential for DDI of ROCTAVIAN with corticosteroids and other small drugs^[24] and such interactions can be predicted using a valid PBPK model. The review and final approval of BEQVEZ included a semi-mechanistic population-PD (popPD) model, which was developed to evaluate FIX activity as supporting evidence for dose justification. During the screening of covariates for their impact on the FIX activity revealed a significant association between age and the rate of synthesis (k_syn) and rate of degradation (k_deg) of the transgene.^[25]

However, many considerations and challenges for AAV GTs exist to apply PBPK M&S approaches to generate evidence for regulatory decision making. Some of these challenges are related to the complex mechanism of action (MoA), limited quantitative PK/PD data for specific disease contexts, variability related to manufacturing attributes and variability in patient treatment response. In recent years, many parameters have been incorporated into MIDD efforts for AAV GTs, which have included; i: considerations about the initial dose/route of administration of the AAV GT, ii: modeling the biodistribution of the vector, iii: estimating the binding and internalization kinetics to different tissues and cell surface components, iv: transgene expression, trafficking and secretion from intercellular vesicles, v: estimating the biodistribution rates and amount of the transgene product in vivo.^[19] Additional consideration include various patient intrinsic or extrinsic factors, AAV tissue tropisms and pre-existing or newly developed neutralizing antibodies against the vector and/or transgene. Given the numerous considerations and mechanisms impacting AAV GT products failure/success, PBPK models may be particularly well suited to integrate the totality of evidence available regarding the biological mechanisms that impact the AAV GT MoA in vivo.

As the underlying biological mechanisms of AAV GTs are further investigated and more thoroughly understood, MIDD efforts may not only continue to generate supportive evidence for drug development efforts but may eventually facilitate the generation of confirmatory evidence.^[26] As discussed in the Agency’s Confirmatory Evidence Guidance,^[26] the FDA considers various levels of studies and approaches including strong mechanistic and PD evidence as sources of confirmatory evidence and PBPK model is well positioned to support mechanistic understanding and generating quantitative PK/PD data for AAV GTs. ^[26] This has the potential to save time and resources that would otherwise be used to designing a dedicated clinical trial to explore different factors that may impact dosing decisions or safety considerations.

Messenger RNA (mRNA) Therapeutics

mRNA products, including vaccines and therapeutics, have emerged as a promising class of biologics, offering a versatile platform for the treatment and prevention of various diseases.^[27] To facilitate the delivery of mRNA molecules to target cells, lipid nanoparticles (LNPs) have become a widely used as effective delivery vehicle, protecting the mRNA from degradation, and promoting its uptake and translation into protein.^{[2, 28]} From therapeutic perspectives, mRNA encapsulated in LNPs (referred throughout this article as mRNA therapeutics) are being evaluated for treatment of several diseases, including cancer,^{[29, 30]} infectious diseases,^{[31, 32]} and rare genetic diseases.^{[33, 34]}

The most common route of administration for mRNA therapeutics is IV administration but other route of administration including inhalation are being investigated in clinical trials.^[35] Following IV administration mRNA therapeutics preferentially distribute to liver, undergo endosomal escape into cytosol and initiate protein synthesis in cytoplasm (Figure 2). This liver tropism is believed to depend on liver physiology (e.g., high blood flow and porous endothelium) and interactions with serum proteins (e.g., apolipoprotein E, APoE) whereby LNPs rapidly bind to APoE that facilitates their targeting to hepatocytes through low density lipoprotein (LDL).^[36]

Figure 2:

Generic minimal PBPK model for LNPs-mRNA and schematic representation of the workflow for developing a PBPK model to predict the biodistribution of lipid nanoparticle (LNP)-encapsulated mRNA therapeutics. The workflow is divided into three main stages: (1) Data Collection, where both LNP-mRNA specific data (e.g., physicochemical and biochemical properties) and system-specific data (e.g., demographic and physiological characteristics) are gathered; (2) PBPK Model Building, which involves defining the model structure, input parameters, and writing ordinary differential equations (ODEs) to describe the system. The minimal model structure includes compartments for plasma, lymph, leaky tissues, and tight tissues, interconnected by lymphatic and blood flow. A detailed intracellular representation shows the LNP-mRNA entering the cell, releasing mRNA, and subsequent translation into transgenic protein. (3) PK/PD Prediction, where the model is verified, validated, and refined using available data. This stage also includes sensitivity analysis and exploration of potential applications of the model. Recognizing the physiological complexities of drug distribution, the incorporation of leaky versus tight tissues compartments in minimal physiologically based pharmacokinetic (mPBPK) models, as supported by the extensive work of William Jusko and colleagues,^{[44, 50]} allows for a more nuanced and accurate representation of transcapillary exchange and drug disposition within various tissues. The model provides time-dependent concentration profiles for mRNA, transgenic proteins, and LNP components across multiple compartments. The representative serum concentration plots are sketched based on single-dose PK data from August et al. 2021.^[47] It is important to note that with multiple dose administrations immunogenicity may lead to changes in a product’s PK/PD profile. Figure created in https://BioRender.com.^[51]

MIDD approaches including PBPK M&S are expected to play an important role for mRNA therapeutics for dosing regimen determination, defining therapeutic threshold and safety and efficacy evaluation. PBPK model-based exposure prediction is especially relevant for mRNA therapeutics due to several factors ^[37]:

1. The need for higher level of protein expression to reach a therapeutic threshold,

2. The requirement for systemic delivery in most cases, involving higher mRNA dose and high delivery efficiency, which may increase risk of toxicity,

3. The necessity for multiple dosing regimen

4. The potential for biological activity and efficacy to be repressed by immunogenicity, particularly in multiple dosing scenarios.

There are some published PBPK models that describe PK and biodistribution of nanoparticles.^[38–43] However, it remains unknown if such PBPK models developed for nanoparticles can be extrapolated or extended to predict PK/exposure of LNPs for delivery of mRNA products. PBPK modeling with minimal compartments that have previously described for disposition of drugs^{[44, 45]} and mAbs^[46] can be developed for mRNA therapeutics to quantitatively describe biodistribution of components of LNPs (e.g. ionizable lipids), mRNA and secreted proteins (Figure 2). In this regard mechanistic PBPK modeling can leverage data on LNPs, mRNA and transgene proteins, including physiochemical characteristics (e.g., size, charge, etc.) and biochemical features (e.g., binding affinity with APoE, enzyme degradation) and physiological information (e.g., organ size, blood flow, and lymphatic flow, etc.). It should be noted that studies have shown each of the three major components of mRNA therapeutics (i.e., LNPs, mRNA and transgene proteins) has their distinct physicochemical and biochemical features that contribute to unique PK profiles.^{[47, 48]} In this regard PBPK models can be developed by leveraging unique features of the components of mRNA therapeutics to predict PK profiles of LNPs, mRNA and transgenic proteins (Figure 2). For example, Miyazawa et al. (2024)^[49] developed mechanistic PBPK-QSP model that simultaneously evaluated the kinetics of LNPs, mRNA and expressed protein.

The PBPK-QSP model developed by Miyazawa et al. (2024)^[49] described the behavior of LNP-based modified mRNA molecules that encode humanized uridine-diphosphate-glucuronosyltransferase (hUGT1A1), a potential treatment for Crigler-Najjar syndrome which prevents proper processing of bilirubin. The study demonstrated the utility of the PBPK-QSP model in optimizing the design of mRNA therapies and predicting PD activity in patients. The key outputs of the PBPK-QSP model include the kinetics of LNPs, mRNA, and expressed enzyme. The PD endpoint was defined as the percentage of patients maintaining a bilirubin level lower than 100 nmol for more than 80% of the treatment time. mRNA stability, translation, and cellular uptake rate were the most critical factors influencing enzyme expression and PD response. To further understand the relationships between model parameters and enzyme exposure, the authors performed sensitivity analyses, which revealed that mRNA degradation rate, translation, and cellular uptake rate were the most sensitive determinants of enzyme AUC. The model also predicted dose-response curves for various dosing schedules and LNP-mRNA design parameters, providing valuable insights into treatment PD effect under different conditions. Although the use of PBPK modeling for mRNA therapeutics is still in its early stages, this study illustrates the potential of PBPK in dose selection and the translation of preclinical findings to exposure-response assessments for predicting human PD responses.

Conclusion/Perspectives

FDA/CBER has seen an increase in regulatory interactions involving PBPK models between 2018 and 2024, indicating growing interest and application in this field. PBPK models have been used to support various aspects of drug development, including dose selection, understanding mechanisms of action, and predicting PK/PD in specific populations.

Opportunities for PBPK modeling for biological products are numerous and promising. PBPK modeling can help optimize dosing regimens, particularly for novel therapies like AAV GTs and mRNA therapeutics. They show potential in supporting pediatric drug development by predicting first-in-pediatric doses while considering developmental factors. PBPK models align with the FDA’s goal to reduce animal testing by leveraging existing data to predict safety, immunogenicity, and pharmacokinetics. They offer a platform for integrating diverse data types, providing a more comprehensive understanding of drug behavior in the body, including complex biological processes involved in gene therapies. Additionally, PBPK models could be valuable in predicting DDIs for new modalities and in estimating drug PK in special populations, such as patients with hepatic impairment, which is particularly relevant for liver-targeted therapies.

In conclusion, the application of PBPK modeling to CGTs faces several challenges. The primary hurdle is the availability of high-quality pre-clinical and clinical data, which is crucial for building and validating these models, especially for rare diseases. The complex MoA of biological products and the high variability in patient responses can be difficult to capture accurately in PBPK models. Manufacturing variability and the potential for immunogenicity add further layers of complexity. From a regulatory perspective, while the use of PBPK modeling is increasing, its acceptance as a primary source of evidence for decision-making is still evolving, particularly for novel therapies. Overcoming these challenges will require collaboration among industry, academia, and regulatory agencies to establish best practices for model development and application in regulatory decision-making.