Advancing FDA New Approach Methodologies from animal models through digital twins

Abstract

In April 2025, the U.S. Food and Drug Administration announced immediate steps toward replacing animal testing for drug evaluation with New Approach Methodologies (NAMs)—modern laboratory techniques mimicking human tissues. However, significant gaps exist between current regulatory frameworks and these technologies’ promise. We argue that specific comprehensive regulatory reforms will improve transition to human-relevant drug-evaluation methodologies, laying groundwork for digital twins, in silico trials, and transformative advances in precision medicine.

Introduction

In April 2025, the United States Food and Drug Administration (FDA) announced immediate steps would be made toward the replacement of animal testing requirements for monoclonal antibodies and other drugs with New Approach Methodologies (NAMs)¹—modern laboratory techniques such as cell-based assays, organ-on-a-chip platforms, and computer-based pharmacological models designed to mimic human tissues and organs (Fig. 1). This announcement largely outlines the agency’s roadmap² for implementing FDA Modernization Act 2.0, a 2022 statute³ that amended the Federal Food, Drug, and Cosmetic Act to remove language explicitly requiring pre-clinical animal testing in drug development and authorized sponsors to use “non-clinical tests” (such as NAMs) to investigate the safety and effectiveness of new drugs prior to testing in humans. Precedent for phasing out animal testing exists in other domains, though with important distinctions from pharmaceutical applications: the European Union has banned animal testing for cosmetics, and the European Medicines Agency (EMA) recently released a roadmap⁴ for eliminating animal testing in chemical safety assessments that emphasized validation frameworks, workforce training, and progress metrics—infrastructure elements that could be adaptable to pharmaceutical NAM implementation—despite significant differences in the complexity of modeling human disease and therapeutic efficacy. While NAM adoption has become a global regulatory priority⁵, the FDA is the first national regulatory agency to promulgate a plan for immediate NAM implementation.

Fig. 1. Evolution of drug development methodologies.

(2D/3D two-dimensional/three-dimensional; AI artificial intelligence; FDA Food and Drug Administration; NAM new approach methodology). Created with BioRender. Parikh, R. (2025) https://BioRender.com/b2f9qvz. In refs.^{14, 22}.

Although the FDA’s announcement largely invokes existing legislative authority³, it nonetheless marks a significant acceleration towards alternative, human-centered methods of evidence generation—laying the groundwork for future integration of computational approaches like digital twins and in silico trials. Just as NAMs aim to reduce the need for drug testing in animals (as well as reduce the risk of unforeseen side effects in patients) through human-relevant alternatives/supplementation, emerging technologies such as digital twins and in silico trials seek to reduce the need for drug testing in humans via virtual, computer-based drug response simulations. To realize this trajectory towards new-technology-mediated advancements in preclinical and clinical drug development, regulatory agencies must urgently address fundamental certification standards, resource allocation challenges, and scientific validation frameworks for NAMs. These approaches will not only determine the success of current NAM implementation (and subsequent effects on drug development and patient safety) but will also lay essential groundwork for next-generation digital twin and in silico trial technologies. Herein, we argue that global regulatory bodies should follow a milestone-based approach to transition from animal testing to NAMs (whether in full or in part) and eventually to digital twins and in silico trials.

From animal models to NAMs: the evolution of drug testing methodologies over the 20th century

The institutionalization of animal testing as a legal requirement in drug development dates back nearly a century. Statutory drug testing in animals was first initiated following the Elixir Sulfanilamide disaster of 1937, wherein a liquid formulation of sulfanilamide—a popular antibiotic deemed safe for use in tablet form but difficult for pediatric patients to swallow—was created by dissolving the medication in diethylene glycol, a poisonous substance otherwise known as antifreeze⁶ (Table 1). This catastrophe led to the death of more than 100 people including many children and prompted the enactment of the Federal Food, Drug, and Cosmetic Act of 1938, mandating safety testing of new drugs in animals prior to human testing and demonstrable human safety prior to marketing⁷. Animal testing became further institutionalized following the thalidomide tragedy of the early 1960s, when a morning sickness medication approved for use in Europe without testing in pregnant animals, caused severe birth defects in over 10,000 children. In response, Congress enacted the Kefauver–Harris Amendments of 1962, requiring drug sponsors to demonstrate drug safety and efficacy before marketing, and incorporate animals (usually two species: one rodent and one non-rodent) in drug metabolism, toxicology, and teratogenicity studies⁷.

Table 1.

Selected historical milestones in animal testing and NAM development

Milestone (Year)	Description
Elixir Sulfanilamide disaster (1937)	• Over 100 people (many children) died after consuming sulfanilamide compounded in poisonous diethylene glycol
FD&C Act (1938)	• Enacted in response to the elixir sulfanilamide disaster • Mandated safety testing of new drugs in animals prior to human testing and demonstrated human safety prior to marketing
Thalidomide crisis (1950s-1960s)	• Thalidomide prescribed to treat morning sickness in pregnant women caused severe birth defects in over 10,000 children • Thalidomide was approved for use in Europe (but not the United States), despite not being tested in pregnant animals
Kefauver–Harris Amendments (1962)	• Enacted in response to the thalidomide crisis • Landmark legislation requiring drug sponsors to demonstrate proof of drug safety and efficacy in humans before marketing, as well as the incorporation of animals (usually two species) in drug metabolism, toxicology, and teratogenicity studies
ICCVAM Act (2000)	• Enacted to coordinate the federal government’s evaluation of alternative testing strategies (particularly those aimed at reducing animal use across federal agencies) • Broadly promotes regulatory acceptance of new/revised scientific testing methods that protect human, animal, and environmental health • Facilitates international collaboration in the development of alternative test methods
Tox21 (2008)	• Federal research collaboration between FDA, NIEHS, NCATS, and EPA • Goal to reduce reliance on animals for toxicity testing and support scientific shift towards NAMs
FDASIA (2012)	• Mandated the FDA’s advancement of regulatory science through the development and adoption of new tools and models for evaluating agency-regulated products
FDA’s Predictive Toxicology Roadmap (2017)	• Culmination of FDASIA • Emphasized FDA’s expansion of internal expertise in non-animal approaches to drug development
FDA Modernization Act 2.0 (2022)	• Removed statutory language requiring pre-clinical animal testing • Authorized sponsors of new drugs to use “non-clinical” tests (i.e., NAMs) to investigate drug safety and effectiveness prior to human testing
FDA Modernization Act 3.0 (2024, first introduction)	• Senate bill passed in Dec 2024 (not yet passed by the House) • Designed to mobilize the FDA’s adoption of Modernization Act 2.0 by detailing actionable steps for implementation

EPA Environmental Protection Agency; FD&C Act Food, Drug, and Cosmetic Act; FDA Food and Drug Administratio; FDASIA FDA Safety and Innovation Act; ICCVAM Act Interagency Coordinating Committee on the Validation of Alternative Methods; NAM nNw Approach Methodology; NCATS National Center for Advancing Translational Sciences; NIEHS National Institute of Environmental Health Sciences; Tox21 Toxicology in the 21st Century Consortium.

However, advances in molecular biology, computational modeling, and high-throughput screening (in vitro) during the 1980s and 90s exposed concerns about the predictive limitations of animal testing for human translation. This is perhaps most clearly highlighted by the 2006 Phase I trial of superagonistic anti-CD28 antibody, TGN1412, which resulted in cytokine storm-induced multi-organ failure in six healthy volunteers—an event attributed to interspecies differences in immune activation between humans and the study’s preclinical animal model^2,8. These and similar concerns center largely on interspecies differences in molecular physiology (which can alter drug absorption, metabolism, and therapeutic/toxicity mechanisms/effects), and pathological etiology. For instance, many animal models in pharmacology rely on artificial methods of induction to mimic disease, which may not fully recapitulate the complexity or natural origins and progression of human pathophysiology. Collectively, these concerns have since driven coordinated efforts across federal agencies—including the National Institutes of Health (NIH)⁹, FDA, and the Environmental Protection Agency (EPA)¹⁰—to support NAM development and reduce reliance on animal testing.

Bridging critical gaps: scientific and regulatory challenges in NAM implementation

As interest in reducing animal use has grown, a diverse collection of NAMs for pharmaceutical testing has emerged—ranging from simple 2D cultures of immortalized human or patient-derived cells in vitro, to complex organ-on-a-chip bioengineering platforms and multi-modal, computer-based predictions of drug behavior (Fig. 1). These computer-based in silico approaches can be sophisticated, integrating vast datasets generated by simpler NAMs, existing clinical data, and artificial intelligence-based analysis. Together, these drug-testing methodologies can offer compelling advantages compared to animal models: direct human-relevance (including cells sourced from human disease), targeted study of specific biological systems, and high-throughput screening—potentially enabling faster, safer, and more cost-effective drug development with better patient response prediction. Yet, significant gaps between the promise of NAMs and the FDA’s current scientific expertise and quality frameworks needed to oversee them risk delaying timely NAM adoption and inhibiting the benefits they are intended to provide.

The most fundamental gap concerns the isolated nature of current NAMs versus the integrated physiology they aim to mimic. While individual NAMs may excel at predicting drug effects or toxicities on isolated biological systems, they cannot model the complex coordination of molecular signaling pathways across the ~50 interconnected tissues and their specialized cells (and chemokines, etc.) routinely evaluated in animal models¹¹. Even integrating multiple NAMs to approximate the function of a single animal model requires substantial resources, yet yields incomplete physiological coverage, as commercial availability remains limited to a narrow range of tissue/organ platforms¹¹. These scientific limitations manifest most clearly in NAMs’ inability to evaluate primary versus secondary therapeutic (and adverse) effects. Many drugs exert initial effects on one organ, only to trigger secondary signaling cascades received by other tissues, which in turn produce the therapeutic or adverse effect. This indirect form of multi-organ communication cannot be captured by most isolated NAM systems, regardless of sophistication. Similarly, NAMs struggle to measure responses involving complex interactions between multiple cell types, growth factors, and tissue remodeling processes manifested by disease conditions they are meant to model. Therefore, while current NAMs can provide valuable human-relevant data for specific biological processes, they have yet to replicate the full complexity of integrated physiological networks. Moreover, NAMs’ use of human cells does not guarantee replication of in vivo biology. Culture conditions often differ from physiological norms (e.g., altered nutrients, tissue-specific oxygen levels, absent mechanical forces), causing phenotypic drift and potential differences in drug responses between patients and NAMs. Thus, clear frameworks for acknowledging/addressing these inherent limitations in regulatory decision-making are necessary.

Such limitations introduce additional regulatory gaps that current frameworks have yet to address. Among the most pressing are gaps in validation standards: what characteristics define a sufficiently validated NAM for regulatory acceptance, and how should agencies certify NAM quality—especially for data generated by in silico platforms integrating multiple sources of data? Equally challenging is the interpretation of conflicting data: when NAM and animal data yield conflicting results, which model should drive regulatory decision-making and in what contexts? Without standardized criteria for NAM fidelity, validation, quality assessment, and conflict resolution, drug sponsors and other members of the drug development/research pipeline face regulatory uncertainty that can stall adoption of these promising technologies. This nascent state of NAM implementation further underscores these challenges: although the FDA has indicated plans to launch pilot programs for drug developers and encourages the inclusion of NAMs data in investigational new drug applications², to the best of our knowledge, no concrete examples of regulatory-grade NAM use—or specific examples of their successes, challenges, or failures—have yet been made publicly available.

Despite these regulatory gaps, the FDA has taken steps to advance NAM adoption through collaborative capacities. For example, the Comprehensive In Vitro Proarrhythmia Assay initiative represents an international effort to support clinical cardiac safety decision-making through in vitro models¹². Similarly, the FDA has engaged in pilot programs evaluating organ-on-a-chip platforms, including Liver-on-a-Chip systems for assessing drug metabolism and hepatotoxicity¹³. However, these efforts remain in developmental and validation stages, serving as proof-of-concept work that has not yet translated into established regulatory pathways for drug evaluation.

The digital twin horizon: from fragmented models to integrated simulations

Given the limitations of both traditional animal models and NAMs, more sophisticated computational approaches are needed, such as digital twins (virtual simulations of individual patients)¹⁴ and in silico trials (computer-based simulations of entire patient populations)¹⁵. These next-generation technologies aim to revolutionize clinical research with the help of artificial intelligence—integrating mathematical lessons learned from NAMs/fundamental experiments and vast genetic, proteomic, pharmacokinetic, and pharmacodynamic datasets, as well as patient records—to predict drug safety and efficacy¹⁶. Unlike current animal models that may not accurately reflect human biology and NAMs that fragment human biology into isolated components, digital twins integrate these elements into comprehensive computer-based avatars of entire patients—capable of simultaneously predicting a drug’s effects across multiple physiologic domains. In silico trials extend this capability by simulating drug performance across entire virtual patient populations, either before or, potentially, as a substitute for clinical testing (assuming validation frameworks can address inherited data quality and fidelity issues)¹⁵. Collectively, these approaches enable personalized predictions of treatment outcomes, including integrated pharmacogenetic, immunologic, and metabolic responses that current testing methods may not adequately capture.

These predictive capabilities position digital twins and in silico trials to fundamentally transform pre-clinical and clinical paradigms by potentially enabling more targeted, efficient, and safe drug evaluation pathways. Pre-clinically, these platforms present opportunities to streamline the translation of laboratory findings to human trials, providing comprehensive safety and efficacy predictions to reduce uncertainty about drug behavior. From a clinical perspective, digital twins and in silico trials could also be used to reduce the number of participants required for Phase 1 and later clinical trials, to optimize drug dosing strategies, and to stratify patients based on predicted risk-benefit responses, even before actual human exposure^15,16. Ultimately, these capabilities could supplement/eliminate the need for extensive human testing by providing robust predictions of therapeutic and adverse effects.

This parallel evolution means that the regulatory frameworks, validation standards, and infrastructural investments that the FDA establishes now for NAM adoption will directly inform the approaches needed to incorporate digital twins and in silico trials into drug evaluation practices of the future. However, successfully navigating these technological evolutions requires a comprehensive regulatory plan—one that addresses both immediate NAM implementation challenges and the foundation for future digital twin integration.

Regulatory roadmap for success: essential pillars for NAM and digital twin implementation

Effective regulatory frameworks for NAMs and digital twins require globally coordinated approaches that address both immediate implementation needs and long-term technological integration. Regulatory approaches that address each technology individually are likely to create fragmented, inefficient processes incapable of accommodating the interconnected nature of these evolving methodologies. Thus, building effective regulatory infrastructure requires organizing solutions along three primary pillars.

Pillar 1: Standardized certification frameworks

Current regulatory approaches to model validation primarily operate on a narrow, individual, indication-specific basis, requiring new platforms to be separately reviewed for each disease or drug application. However, this fragmented approach risks reinforcing the very inefficiencies that NAMs and digital twins aim to overcome, while missing critical overlaps between pathophysiologic conditions. For example, cardiac arrhythmia modeling requires both electrophysiological and ischemic modeling components since many arrhythmias result from ischemia—and approving them separately loses critical physiological integration/complexity/accuracy. Ultimately, narrowly-scoped, piecemeal acceptance of individual models could discourage comprehensive platform development, reinforcing siloed, single-purpose tools lacking the integrated potential needed for more advanced—and accurate—drug response predictions. Thus, regulatory agencies should adopt platform-based certification, allowing a single NAM or digital twin model to be validated across multiple therapeutic contexts once it has demonstrated core performance standards (such as predictive accuracy with known human responses). This approach would also enable the enactment of standardized criteria defining when such drug evaluation technology platforms can substitute for traditional testing methods. Implementing these standards early would ensure consistent and efficient evaluation of both current and next-generation technologies, with full recognition of their potential to supplement or perhaps replace traditional animal and human testing approaches.

Pillar 2: Strategic resource investment and workforce development

Successful implementation of NAMs—and subsequently digital twins—also requires addressing critical resource and infrastructure gaps. The transition period from animal-only testing to integrated NAM adoption requires that drug sponsors and regulatory reviewers simultaneously evaluate both methodologies to determine when NAMs can appropriately be implemented across different therapeutic areas/indications. Not only will this dual evaluation increase workloads and acute costs for sponsors and reviewers¹⁷, inadequate initial investment could extend this expensive transition period beyond what is necessary. Thus, effective NAM implementation requires coordinated investment across several major areas: workforce pipeline development through strengthened academic training programs and university research initiatives; enhanced NAM training for current regulatory reviewers; and infrastructure development for international NAM-based data repositories (such as server clusters and secure sharing frameworks), including establishing industry data-sharing partnerships¹⁸ to facilitate model validation.

Realizing these investments will require structured collaboration across biomedical sectors. For instance, academic researchers may be unaware of the FDA’s specific evidentiary priorities until late in NAM development; thus, greater collaboration between the FDA and NIH—including FDA-guided requirements and milestones for NIH-funded university research—could align stakeholders from the outset and accelerate NAM progress. Industry organizations such as BIO and PhRMA could also encourage pre-competitive NAM data-sharing among member companies, building on models like TransCelerate BioPharma Inc’s initiatives for pharmaceutical collaboration. Upcoming Prescription Drug User Fee Act (PDUFA) negotiations also present an opportunity to allocate user fees toward specialized infrastructure and regulatory reviewer training required for NAM advancement. Together, these mechanisms could distribute resource burdens while accelerating the evidence base needed for regulatory acceptance of NAMs.

Pillar 3: Adaptive regulatory architecture

Finally, regulatory frameworks must also accommodate rapid technological advancement of drug evaluation platforms through flexible, iterative validation approaches able to evolve alongside emerging platform capabilities without requiring complete regulatory overhauls. This adaptive architecture requires two key components: controlled testing environments and clear decision-making protocols. The first component involves regulatory sandboxes—structured environments where new technologies can be tested under FDA supervision with fit-for-purpose regulatory requirements. One such example of this approach is the FDA’s Software Pre-certification Pilot Program¹⁹. Rather than requiring pre-market review for each software update in a medical device—a process incompatible with software’s rapid/continual updates—the Program granted streamlined pathways for lower-risk modifications while shifting to continuous post-market surveillance (for pre-qualified companies). This model demonstrates how regulatory oversight can transition from intensive upfront review to real-time monitoring. Applied to NAMs and digital twins, such sandboxes would enable controlled testing of novel drug evaluation platforms while generating evidence to inform sponsors and evolving regulatory guidelines through systematic data collection.

Equally important are clear decision-making frameworks that can address emerging scenarios while maintaining safety standards, ensuring regulatory pathways can accommodate unforeseen technological developments and novel applications. For instance, agencies need pre-established criteria for determining when novel NAM technologies can substitute for traditional testing methods—such as decision trees that evaluate platform validation status, data quality metrics, and biological relevance thresholds. Similarly, frameworks should define how to handle hybrid approaches that combine multiple methodologies—establishing clear protocols for weighing different types of evidence and determining acceptable risk-benefit profiles for regulatory approval.

Global coordination and regulatory collaboration

Standardized certification frameworks, enhanced data sharing infrastructure, and adaptive regulatory architectures represent challenges shared across regulatory agencies worldwide. Recent work by the EMA has identified similar barriers to NAM adoption, including limited data sharing between developers and regulators, unclear terminology and validation requirements, and insufficient funding to support regulatory evidence generation^20,21.

International coordination will be essential to ensure that regulatory advances in one jurisdiction can be leveraged globally, avoiding duplicative validation efforts and accelerating adoption of human-relevant testing approaches. Forums such as the International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH)—which brings together regulatory authorities and pharmaceutical industry representatives to develop unified technical guidelines—provide established mechanisms for such coordination. Similarly, working groups dedicated to implementing the 3Rs principles (principles guiding the Replacement, Reduction, and Refinement of animal use) offer venues for sharing validation strategies and harmonizing acceptance criteria across jurisdictions. Thus, by establishing clear regulatory frameworks and articulating pathways toward next-generation computational approaches, regulatory agencies can collectively advance the transition from animal testing through NAMs to digital twin and in silico trial technologies.

Building tomorrow’s drug development ecosystem

The convergence of NAMs and digital twin technologies represents a pivotal moment in drug development regulation, where the frameworks established today will determine whether we achieve incremental improvement or more significant advances in how new drugs are assessed and how quickly they reach patients in need. Collectively, the pillars proposed herein represent more than incremental policy adjustments—they constitute the foundations for a modernized drug development ecosystem that prioritizes human-relevant testing methodologies and personalized therapeutic approaches. Implementing these frameworks positions regulatory agencies to lead a transition that extends beyond reducing animal testing to encompass meaningful advances in precision medicine and reduce uncertainty in clinical trial design and drug behaviors.

The implications for pharmaceutical innovation are profound. Companies investing in NAM and digital twin technologies today will benefit from clear regulatory pathways that reduce developmental uncertainty and enable more efficient resource allocation toward promising therapeutic candidates. Simultaneously, patients will benefit from improved access to innovative treatments and reduced exposure to potentially ineffective or harmful drugs through more precise drug-response predictions. However, the window for establishing regulatory foundations is narrowing as these technologies rapidly evolve. Further, establishing these foundations now avoids the more complex task of later harmonizing divergent, fragmented practices. Thus, proactive regulatory leadership that anticipates technological evolution and establishes robust validation standards now will determine whether the FDA achieves its intended goals to reduce animal reliance in drug development and improve developmental efficiency in a timely manner.

Realizing the full potential of human-relevant drug testing requires sustained commitment to addressing fundamental knowledge gaps and building regulatory infrastructure that will support decades of continued technological advancement. By establishing robust validation frameworks, investing in expertise/education, and creating adaptive pathways today, we can ensure that NAMs and digital twins deliver tangible benefits for pharmaceutical innovation and patient outcomes.

Author contributions

All authors (A.L.E., H.F.L., N.S., R.B.P.) conceived, drafted, critically revised the article, and provided final approval.

Data availability

No datasets were generated or analysed during the current study.

Competing interests

The authors declare no competing interests.