Modeling Approaches for Determination of Pharmacokinetics/Pharmacodynamics

Syringe aimed at the center of a target. Pharmacogenetics concept, symbolizing precise drug delivery based on genetic profiles. Low poly style | Image Credit: ©Елена Бутусова -stock.adobe.com

Successful drug development requires in-depth understanding of the pharmacokinetic (PK) and pharmacodynamic (PD) properties of drug substances and their behavior once administered to patients. PK/PD modeling is a mathematical approach toquantifying the absorption, distribution, metabolism, and excretion of a compound within the body (PK), the effects of a drug substance over time (PD), and the relationship between them (1). It has many applications from very early-phase drug development (target and lead candidate selection), through preclinical evaluation (quantitative characterization) to clinical studies (trial designs and optimization of dosing regimens), supporting more-informed decision making through a project’s lifetime (2).

The use of PK/PD modeling techniques during the development of biologic drugs has expanded noticeably as computational techniques have advanced, cost-effective computing power has increased, and the complexity and diversity of modalities and drug delivery systems in the pipeline has grown.

Rising biologics complexity and diversity drive modeling use

Biologics, observes Iordanis Kesisoglou, PK/PD modeler, metabolism at Labcorp, often exhibit non-linear pharmacokinetics and complex mechanisms of action that go against the usual assumption of linearity that study designs take. “PK/PD modeling is essential for describing non-linear kinetics and predicting efficacy and safety for these types of biologics,” he says. In addition, Kesisoglou notes that with the increasing complexity of modalities under development today, greater use of PK/PD modeling supports more precise understanding of drug behavior, optimization of dosing regimens, and reduced need for extensive clinical trials.

The rise in complexity and diversity of biologic drugs, agrees Hannah M. Jones, senior vice president and head of Certara’s Simcyp PBPK Modeling Services, has necessitated advanced biosimulation tools like quantitative systems pharmacology (QSP) and physiologically-based pharmacokinetic (PBPK) modeling. “These models enable the simulation of complex physiological interactions, such as drug target binding, clearance mechanisms, and immunogenicity. In addition, these platforms allow researchers to predict human outcomes with better accuracy, reducing reliance on traditional trial-and-error methods and enabling data-driven decisions from early discovery to clinical stages,” she explains.

Informing and streamlining biopharmaceutical drug development

There are, in fact, several benefits to PK/PD modeling, according to Rhylee Decrane, PK/PD modeler, metabolism with Labcorp. They include informed decision-making, reduced development costs, and enhanced understanding of drug mechanisms.Piet van der Graaf, senior vice president and head of quantitative systems, pharmacology at Certara, also notes that modeling and simulation approaches streamline drug development by predicting exposure, efficacy, safety, and dosing.

The particular benefits of PK/PD modeling, van der Graaf continues, vary based on the model. Mechanistic models simulate biological processes, offering insights like target-mediated drug disposition, while population-specific models evaluate variability across diverse patient groups. For biologics, he comments, PBPK and QSP models integrate physiological and molecular data, optimizing predictions across drug modalities and development stages, from preclinical to post-marketing.

Different types of modeling (e.g., empirical, mechanistic, population-based) also offer unique advantages depending on the development stage and modality, Decrane adds, allowing adaptability for when new data are presented from completed studies. “PK/PD modeling can be employed in early stages of development to help select drug candidates with favorable PK endpoints. This methodology can also be used in the translation of pre-clinical trials to help support scaling of different species. The predicted data generated from simulated scenarios can, meanwhile, be used to optimize first-in-human dose selection, predict efficacy and safety, analyze drug-drug interactions, and support regulatory submissions,” she says.

Several modeling approaches

There are two main categories of PK/PD modeling strategies according to Jones: mechanistic and population-based. The former, she explains, comprise elaborate physiological frameworks simulating biological and biochemical pathways, while the latter are data-driven approaches to explore pharmacokinetic and pharmacodynamic variations within specific patient subgroups.

In addition to these main types of models, Kesisoglou also highlights machine learning-based PK/PD models as being valuable. “Each approach has its strengths and is chosen based on the specific application and available data,” he says. All the approaches, he comments, support model-informed drug development (MIDD), which is receiving increasing attention and encouragement by regulatory bodies such as FDA in the United States and the European Medicines Agency in Europe.

“By leveraging MIDD approaches, pharmaceutical companies are positioned for higher success rates and faster innovation cycles for biologics,” says Jones. She highlights mechanistic PBPK and QSP modeling for their ability to not only enhance drug development efficiency, but also align with ethical trends, such as minimizing animal studies.

Many factors determine the best approach

The choice of modeling approach is made taking several different aspects into consideration, including the research question/study objective, drug modality, available data, and the stage of development.

Mechanistic models, for instance, are often used in early discovery to understand drug mechanisms, while population models are used in clinical development to optimize dosing regimens based on variability of a particular population, according to Decrane. In addition, mechanistic models are commonly used for biologics due to their complex mechanisms of action, while empirical models are often used for small molecules with well-characterized PK/PD relationships, observes Kesisoglou.

As an example, van der Graaf points to the use ofmechanistic PK/PD or QSP models for predicting interactions like immunogenicity for biologics with complex pharmacological mechanisms. “Going a bit deeper,” he observes, “leveraging QSP modeling as an immunogenicity prediction tool enables drug developers to optimize critical parameters such as dose, frequency, and patient population to mitigate risks and enhance biologics development.”

Jones, meanwhile, notes that mechanistic models are particularly useful for mechanistic pharmacological target engagement and clearance prediction for monoclonal antibodies. Lastly, Jones comments that QSP models excel in applications where detailed mechanistic insights are lacking, as this technique enables scientists to simulate clinical trial scenarios that are otherwise prohibitively expensive or impractical to test experimentally.

Complexity creates challenges too

While the complexity of biologic modalities has driven greater use of PK/PD modeling in drug development, it has also created challenges to its use. “One of the biggest challenges to effective use of PK/PD modeling for biopharmaceutical drug candidates is capturing the complexity of biologic interactions, such as multi-faceted clearance or immunogenicity,” van der Graaf states. Integration of data from diverse sources (in vitro, animal, clinical) and regulatory shifts toward modeling-based submissions are additional difficulties he highlights.

Other important issues include, according to Decrane, limited data availability and the need for specialized expertise. Mechanistic models (e.g., QSP and target-mediated drug disposition [TMDD]), she explains, can be harder to use due to their complexity, while empirical models may be limited by data quality.

Different modalities also present different hurdles to achieving effective PK/PD modeling. Gene therapies, for instance, present greater challenges due to their novel mechanisms of action, while antibody-drug conjugates (ADCs) involve multiple components, each with distinct PK and PD aspects that must be considered, Decrane comments.“As more novel mechanisms are added into a model, the risk of overfitting increases and must be addressed,” she says.

Increasing role for advanced algorithms

Incorporation of artificial intelligence (AI) and machine learning (ML) capabilities into PK/PD models has helped drug developers overcome some of these challenges. AI and ML have, observes Kesisoglou, improved PK/PD modeling by enhancing model development and validation, leading to more robust and predictive models that are able to analyze large datasets, identify complex patterns, and predict outcomes with greater accuracy.

AI and ML have also, according to Jones, accelerated the data simulation and pattern recognition capabilities of all types of PK/PD models. In addition, she notes that ML tools have proved useful for automation of the assessment of model goodness of fit, robustness, and parsimony. One example noted Jones is the use of ML to link algorithms in QSP models to clinical endpoints, allowing for the prediction of clinical efficacy ahead of actual trials.

Start early and stay up-to-date to optimize modeling benefits

Modeling experts emphasize the importance of integrating PK/PD modeling early in the development process. Decrane highlights using a combination of mechanistic and population-based approaches and leveraging advanced technologies like AI/ML. QSP and PBPK modeling for mechanistic biosimulation should be implemented with the goal of realizing scalable insights, according to van der Graaf.

“Developers with the goal of increasing useful data generation should first conduct gap analyses to align modeling strategies with scientific goals and regulatory expectations,” van der Graaf explains. “From there,” he continues, “they can use mechanistic modeling tools for comprehensive predictions across preclinical and clinical boundaries and proactively engage regulatory bodies early to validate modeling approaches.”

Model development, adds Decrane, is an iterative process that should include integration of new data as and when it becomes available to inform and improve existing models. “By creating a model that can adapt to changes in the study, a better-informed decision can be made for the drug candidate,” she says. Decrane also believes that biologics developers can increase the likelihood of obtaining useful data by ensuring high-quality data collection, collaborating with experts, and continuously validating and refining models.

It is also important to stay abreast of PK/PD modeling developments, according to Kesisoglou, because the field is dynamic and evolving, with ongoing advances in technology and methodology. “Staying up-to-date with the latest developments and best practices is crucial for successful application in biologics development,” he contends. He also points out that the adaptive design of PK/PD models will continue to allow success in the future as biologics and other compounds evolve.

References

1. Zou, H.; Banerjee, P.; Leung, S.S.Y; and Yan, X. Application of Pharmacokinetic-Pharmacodynamic Modeling in Drug Delivery: Development and Challenges.Front. Pharmacol., Sec. Transl. Pharmacol,2020 11. DOI: 10.3389/fphar.2020.00997
2. Penney, M. and Agoram, B. At the Bench: The Key Role of PK–PD Modelling in Enabling the Early Discovery of Biologic Therapies.Br. J. Clin. Pharmacol.2014 77(5):740–745 DOI: 10.1111/bcp.12225

About the author

Cynthia A. Challener, PhD, is a contributing editor to BioPharm International®.