The pharmaceutical industry is undergoing a paradigm shift, moving towards New Approach Methodologies (NAMs) to reduce reliance on traditional animal models. At the forefront of this shift is Physiologically Based Pharmacokinetic (PBPK) modeling, a mechanism-driven computational approach recognized by global regulators, including the U.S. Food and Drug Administration (FDA) and the European Medicines Agency (EMA), as a pivotal tool for drug safety and dosing decisions.
PBPK modeling effectively creates a “digital human,” mechanistically simulating how a medicine behaves within the body. This predictive framework integrates two primary sets of data:
1. System Parameters (Physiology): These define the body as a network of physiological compartments (e.g., liver, kidney, brain) connected by blood flow. Crucially, for studying vulnerable groups, PBPK models incorporate ontogeny functions to account for age-dependent developmental changes in physiology, drug-metabolizing enzymes (DMEs), and drug transporters. This is particularly advantageous over traditional scaling methods which ignore these age dependencies.
2. Drug Parameters: This includes the drug’s specific physicochemical properties (e.g., solubility, permeability) and its interactions with metabolic pathways.
By solving mathematical equations based on these inputs, PBPK predicts quantitative systemic and tissue-specific drug exposure, offering insights into concentrations at target sites that are unavailable through conventional methods.
PBPK modeling has proven invaluable across multiple critical phases of drug development:
• Drug-Drug Interactions (DDI): This is the most frequent regulatory application, constituting 81.9% of all PBPK submissions reviewed by the FDA between 2020 and 2024. Models dynamically predict the impact of enzyme or transporter inhibition on drug exposure, informing clinical risk management and sometimes waiving the need for extensive clinical DDI studies.
• Dose Extrapolation in Special Populations: PBPK is crucial for determining dose recommendations for patients with organ impairment (7.0% of instances) and for pediatric populations (2.6% of instances).
◦ Case Examples: PBPK modeling successfully supported the approval of Rivaroxaban by validating body weight-adjusted dosing regimens and extrapolating exposure to untested pediatric age groups. Similarly, a model was utilized to address regulatory questions about dose confirmation for nilotinib in sparse-data populations (2 to <6-year-olds), supporting the drug label extension.
Regulatory Scrutiny and Credibility Requirements
Regulatory bodies operate under strict requirements, demanding that any PBPK model supporting a submission establish a complete and credible chain of evidence linking in vitro parameters directly to clinical predictions.
Regulators assess models based on three core criteria:
1. Reliable Parameters: Input values must be derived from robust experiments rather than unverified assumptions.
2. Direct Clinical Verification: Predictions must be confirmed against relevant clinical data.
3. Complete Chain of Evidence: The model must connect in vitro data, calibration, and prospective prediction seamlessly.
Models are categorized by regulatory reviewers based on adherence to these principles:
• Adequate: Sufficiently verified model supporting a regulatory decision (e.g., LIVDELZI DDI prediction).
• Adequate with Limitations: Model is useful but contains caveats, such as reliance on empirical calibration rather than a full mechanistic explanation (e.g., LEQSELVI).
• Inadequate: Model contains substantial deficiencies, such as using non-mechanistic adjustments or unverified assumptions, rendering it insufficient to support the proposed decision (e.g., ATTRUBY).
Key Challenges and Future Directions
Despite its growing adoption, PBPK modeling faces several limitations that constrain its broader application:
• Parameter Uncertainty: There is a lack of consensus in the literature regarding key developmental parameters, particularly the ontogeny profiles for DMEs like CYP3A4, which introduces uncertainty in predictions, especially in neonates.
• Structural Oversimplification: The reliance on empirical adjustments or “curve-fitting” to match observed data, rather than true mechanistic modeling, compromises predictive power for untested scenarios.
• Data Scarcity: High-quality clinical data in the most vulnerable populations (neonates and infants) needed for robust validation is scarce.
• Lack of Standardization: The field currently lacks internationally harmonized, quantitative validation criteria.
Looking forward, the evolution of PBPK modeling involves combining it with advanced computational technologies. Integrating PBPK with Artificial Intelligence (AI) and Machine Learning (ML) algorithms is expected to significantly enhance parameter prediction efficiency, optimize model structures, and improve overall predictive precision, laying the foundation for unified verification standards and personalized medicine strategies. Hybrid approaches also involve coupling PBPK with other models, such as Population PK and Pharmacodynamic models, to simulate tissue-specific drug exposure more accurately
Reference: