New Approach Methodologies: What Clinical Pharmacologists Should Prepare For

Introduction

As New Approach Methodologies (NAMs) gain regulatory momentum for Investigational New Drug (IND) applications, clinical pharmacologists are uniquely positioned to collaborate with preclinical and translational teams in the development, quantification, and implementation of these methodologies. Their involvement is critical to ensuring a reduction in animal use to support early-phase drug development.

Regulatory Shifts and the Role of NAMs

Preclinical evaluations of drug pharmacology, toxicity, and pharmacokinetics have traditionally relied on in vivo animal models. However, species-specific differences in drug target homology, pathophysiology, and immune system function often limit the translational relevance of these models for predicting clinical outcomes. These limitations are especially pronounced for candidate drugs designed to interact with human-specific targets that are not expressed in animals, lack sufficient homology, or exhibit distinct pharmacological effects, ultimately leading to questionable translational value of traditional animal models¹.

Nonhuman primates (NHPs) are generally considered more translatable to humans than small animal models for evaluating immunotherapies, but several cases have demonstrated limitations to NHPs. The tragic TGN1412 incident and the inconsistent findings from preclinical studies of checkpoint inhibitors exemplify these challenges^2,3. For example, ipilimumab showed minimal histological safety concerns in NHP studies, yet it has exhibited the highest incidence of immune-related adverse events (irAEs) among the first few approved immunotherapies. In contrast, pembrolizumab raised early safety concerns in NHPs, such as increased spleen weight and focal mononuclear cellular infiltration, but ultimately demonstrated a relatively favorable safety profile in clinical settings.

Recognizing these limitations, the U.S. Food and Drug Administration (FDA) has recently introduced opportunities to waive certain animal testing requirements, especially for antibody therapeutics using NAMs⁴. NAMs broadly refer to in vitro, in silico, or combination of both that can be used to reduce, refine, and replace animal studies in research and drug development (Figure 1). Although not entirely new in the field of immunotherapy, NAMs represent a pivotal shift in how drug candidates are evaluated during IND preparation. Several agents, such as bispecific T cell engagers, have already advanced to clinical trials supported by in vitro functional assays rather than traditional in vivo animal pharmacology and toxicity studies.

Figure 1.

Types of New Approach Methodologies (NAMs), Context-of-Use, Qualification, and Clinical Translation. Mechanistic models, such as physiologically based pharmacokinetic (PBPK) models and quantitative systems pharmacology (QSP) models, are essential for contextualizing NAM outputs within clinical settings and enabling meaningful translation to human outcomes. The qualification and regulatory acceptance of NAMs often require the integration of artificial intelligence and machine learning (AI/ML) tools to establish connections between NAM-derived readouts and clinical datasets. In addition to clinical datasets, biological reference data, such as those from the Human Protein Atlas and Human Immune Atlas, are also critical for generating supporting evidence for NAM development and quantification.

With NAMs now officially recognized in regulatory review, it is timely to reflect on how clinical pharmacologists can best prepare for this paradigm shift to support IND applications and early-phase clinical trial designs.

Major Roles of Clinical Pharmacologists in the NAM Era

Most preclinical evaluations using animal models follow standardized protocols that typically do not require much input from clinical pharmacologists. As a result, NAMs have been viewed as the primary responsibility of preclinical and translational scientists, with limited involvement from clinical pharmacologists, whose primary focus is often on clinical trial design, execution, and interpretation. However, unlike conventional animal models, NAMs present both challenges and opportunities that require close collaboration across preclinical, translational, and clinical disciplines. The development, validation, and regulatory qualifications of NAMs should extend well beyond the preclinical domain. As clinical pharmacologists, we have an increasingly critical role to play in guiding how NAMs are integrated into the broader drug development process.

This raises a key question: How should clinical pharmacologists prepare for the integration of NAMs into modern drug development? Drawing from our ongoing experience, we offer the following perspectives:

1. Defining the Context-of-Use: A Core Responsibility

Recent progress in the fields of tissue bioengineering and stem cell technology have enabled the development of cell-based models that can recapitulate various aspects of human physiology and pathophysiology. These innovations have given rise to advances in vitro platforms such as 3D cell cultures, organoids, and organ-on-chip systems, all of which fall under the umbrella of in vitro NAMs^5,6. These platforms hold substantial promises for improving the prediction of drug pharmacology, toxicity, and safety by providing more physiologically relevant human models compared to traditional animal studies.

Despite their scientific potential, many current in vitro NAMs still play a limited role in generating IND-enabling evidence. This limitation is primarily due to challenges related to model complexity, inter-laboratory variability, and a lack of standardized regulatory qualifications. Furthermore, NAM development is often led by engineers seeking to improve physiologic or pathophysiologic mimicry and are thus not designed to achieve endpoints relevant to clinical pharmacologists. As in vitro NAM technologies become increasingly complex, so too does the variability in their performance, making consistent interpretation and cross-site reproducibility difficult. Furthermore, most NAMs have not yet been validated or qualified through regulatory pathways such as the FDA’s Drug Development Tool (DDT) qualification program or ISTAND initiative⁷. Thus, collaboration in early model development among clinical pharmacologists and NAM engineers is essential for developing systems that strike a useful balance between model complexity and clinically relevant outcomes and datasets. Currently, the FDA has begun formally recognizing NAMs in specific contexts, defining the contexts-of-use for in vitro NAMs thus becoming one of the critical responsibilities.

On the other hand, much simpler models with specific, well-defined COU have been widely adopted in many IND submissions. For example, fit-for-purpose 2D in vitro tumor and T cell co-culture systems, such as bispecific T-cell engager-induced cytotoxicity assays, have been accepted under ICH S6(R1) guidance. These systems are explicitly designed to evaluate clearly defined pharmacologic or toxicologic endpoints and have been successfully used to support the first-in-human (FIH) dose selection of immunotherapies. This experience highlights that regulatory compliant NAMs must have a clearly defined COU, one that strikes a balance between technical complexity and clinical interpretability. Furthermore, this well-defined COU serves to define NAM engineering design parameters. Without reproducible protocols and a well-defined COU, regulatory agencies are unlikely to accept data from NAMs as stand-alone evidence for safety or efficacy.

To define such COUs effectively, clinical pharmacologists must be directly involved in NAM design and development. Early collaboration with NAM developers and preclinical or translational teams ensures that experimental designs are aligned with clinical objectives, relevant patient populations, and intended therapeutic applications. This type of interdisciplinary coordination is essential for generating interpretable, regulatory-grade data that supports IND submissions and informs early clinical decision-making.

2. Leveraging Artificial Intelligence and Machine Learning (AI/ML) to Support NAM Qualification and Translation

Advanced NAMs are expected to yield human physiologically relevant outcomes that support clinical translation. However, many deep phenotypic readouts in NAMs, such as transcriptomic changes and spatial imaging features following drug exposure, are frequently difficult to directly correlate with clinical outcomes without AI/ML tools. We believe a comparative, class-based approach can enhance the translational utility of these deep phenotypes derived from NAMs. By anchoring NAM-derived findings to known agents within the same therapeutic class, researchers can better assess the clinical relevance of these phenotypes. In this context, clinical pharmacologists, working closely with AI/ML experts, can leverage large-scale drug development datasets to define within-class benchmarks and support the regulatory qualification of NAMs. AI/ML models can then be applied to translate high-dimensional NAM phenotypic data into clinically meaningful predictions. This strategy is particularly valuable for next-in-class agents, where established comparators exist. Many early-phase development decisions, such as selecting a FIH dose or evaluating the benefit-risk profile, are inherently comparative. Early involvement of clinical pharmacologists can guide the identification of appropriate comparator compounds, interpretation of differential responses, and contextualization of NAM findings within clinically relevant parameters.

For first-in-class agents, where historical comparators are lacking, a broader, weight-of-evidence approach is often necessary. This includes integrating target biology, genomic profiling, systems pharmacology, Mendelian randomization, and causal inference techniques to assess the safety liabilities of pharmacological targets and drug candidates. Here, strong collaboration among clinical pharmacologists, data scientists, and AI/ML experts is essential. Publicly available datasets, such as the Human Protein Atlas and the Human Immune Atlas, can be mined to evaluate tissue-specific expression of pharmacological target, biological function, and potential safety concerns. These data can be effectively leveraged using AI/ML tools to generate evidence that complements NAMs in a weight-of-evidence framework, enabling early risk assessment, dose selection, and prioritization of candidate molecules. When substantial uncertainty persists, adaptive clinical strategies, such as Phase 0 (microdosing) studies, may be warranted to characterize pharmacokinetics and early safety prior to advancing to full-scale trials⁸.

3. Expanding Mechanistic Models to Integrate in vitro NAMs

A key strength of in vitro NAMs lies in their ability to deliver human-relevant mechanistic data, especially in biological contexts that are difficult to replicate in traditional animal models. NAMs can be used to explore drug effects and safety under a wide range of physiological conditions, including those representing vulnerable subpopulations. Compared to in vivo animal models, NAMs enable mechanistic evaluations of cellular and molecular responses under controlled, human-specific conditions—supporting investigations that help uncover potential sources of clinical heterogeneity.

However, while NAMs yield rich mechanistic data, their experimental readouts often do not directly inform early clinical trial decisions, such as dose selection or safety margins. To bridge this translational gap, NAMs can be integrated with mechanistic modeling tools, such as quantitative systems pharmacology (QSP) and physiologically based pharmacokinetic (PBPK) models, which are core components of the clinical pharmacology discipline. These models help translate NAM-derived mechanistic findings into clinically relevant predictions, enabling more robust and informed decision-making in early drug development^9,10. On the other hand, mechanistic modeling approaches such as PBPK, QSP, and systems biology can play a critical role in the design and optimization of NAM studies themselves. By simulating expected pharmacokinetic and pharmacodynamic profiles across a range of physiological conditions, these models can help identify key parameters to measure, optimal sampling time points, and critical biological pathways to interrogate.

There is a growing trend to integrate mechanistic models with AI/ML approaches, as they can be highly complementary and synergistic. This integration is particularly valuable in situations where datasets are limited for training advanced AI/ML models. In such cases, mechanistic models can provide a pharmacological framework to inform the selection of critical features and guide the design of neural network architectures, an approach often referred to as mechanistic learning models. Conversely, AI/ML approaches can help distinguish signal from noise in biological data, reduce data dimensionality, and automate the comparison of alternative mechanistic models, thereby facilitating the development of AI/ML-enhanced mechanistic models. The integration of both approaches can become common for NAM developments and qualifications.

Clinical pharmacologists are uniquely positioned to lead this integration. Clinical pharmacologists bring a unique set of translational competencies that are critical in NAM development and qualification. For example, PBPK models can be used to extrapolate drug disposition from multiple organ-on-chip systems to predict human pharmacokinetics and drug-drug interaction in pharmacokinetics, while QSP models can translate in vitro NAM efficacy or toxicity data into predictions of clinical exposures, thereby informing FIH dose selection strategies. Moreover, clinical pharmacologists are highly experienced in defining exposure-response (E-R) relationships, performing model-informed dose selection, and evaluating variability across patient populations. Their training in bridging pharmacokinetics, pharmacodynamics, and regulatory strategy positions them to ensure NAM-derived data are clinically interpretable and usable in trial design and IND submissions.

Outlook

As NAMs gain regulatory momentum and technological maturity, clinical pharmacologists should adapt and expand their roles to embrace this paradigm shift. We are uniquely positioned to ensure that in vitro NAM-derived insights are contextually relevant, mechanistically grounded, and integrated with modeling frameworks to support IND submission and inform early clinical development. This includes defining clinically meaningful endpoints, specifying exposure ranges, and establishing comparators where possible. Furthermore, collaboration with AI/ML scientists, NAM engineers, toxicologists, and data analysts will be critical to unlocking the full potential of NAMs for IND-enabling applications.

Ultimately, our ability to guide cautious and informed early-phase trial design, grounded in mechanistically and clinically relevant NAM data, will be central to reducing reliance on animal models, while maintaining our commitment to patient safety and scientific rigor.