Rethinking Safety Evaluations: From Animal Models to NAMs

We need fit‑for‑purpose research models and to filter the marketing illusion that NAMs are ready to go

New Approach Methodologies, or NAMS, have been gaining momentum in the pharmaceutical industry, particularly for their use in the safety assessment of medicines. Regulators in both the US and Europe are enabling this change. In a continuous effort to support the 3Rs (replacement, reduction, and refinement of research animals), the US passed the FDA Modernization Act 2.0 in 2022, which clarified that non-animal alternatives can support first-in-human Investigational New Drug (IND) applications. In parallel, EU and US regulators are issuing a new guidance documents (see references below) supporting how NAMs could be integrated into a safety assessment framework to help address the pace and scope of NAMs development.

With that said, the current NAMs publicity can be misleading. It implies the industry is ready with a line-up of alternative methods, when the reality is slightly different for those of us on the frontlines of research. Whilst regulators are actively discussing roadmaps for implementing NAMs, the technology has yet to be defined, selected, or validated. The first step toward this will likely be two-tiered, with NAMs used alongside current in vivo physiological systems and as part of a weight-of-evidence strategy. While there are already cases in which in vitro assessments alone can be used, they are currently in the minority and specific to a modality or indication. An adjustment to the language we use may also be useful.  As noted in a recent European Medicines Agency reflection paper, regulators referred to ‘using 3Rs approaches’ rather than positioning NAMs as direct substitutes. The Medicines and Healthcare products Regulatory Agency (MHRA) captures this distinctly ‘We recognise that asking alternative methods to replicate what is currently seen in animal studies is often not the right question. The right question is whether the alternative methods give information necessary to make the appropriate decision in relation to biology, efficacy, safety, or toxicity’.

Safety assessments are multidimensional, and drug development is highly regulated. The key principles are outlined in several regulatory guideline documents, and the regulators uphold these standards. The FDA recently noted in its NAMs guidance document that  ‘Although federal laws and regulations do not stipulate exactly which nonclinical tests sponsors must conduct nor in what model, a common set of studies (e.g., general toxicology, primary pharmacology, and safety pharmacology) is generally conducted before clinical trials in humans are initiated.’
For someone who has witnessed and submitted hundreds of regulatory documents, the guidelines are quoted, and if they are not followed or explained clearly, the dossier will be questioned. In the USA, the consequence can mean being put on clinical hold, in other words, the clinical trial won’t be starting
It is going to take a lot of work to find the holy grail for replacing all in vivo animal studies with in vitro tests alone. Pursuit of this goal is clearly underway. However, the current roadmap is meandering with plenty of diversions and dead ends. It is a challenge to make sense of the numerous options offered, which today are far from being accepted and validated from a regulatory perspective. Moreover, the guidelines have not yet been updated.

Roadmap and Context of Use

Developing any NAM for use in a safety assessment needs clarity on its context of use. It will be important to understand how a NAM designed to address hazard identification fits into a current drug’s derisking and safety assessment plan. We need to select the model that is fit for purpose and filter the marketing illusion that NAMs are ready to go.

The safety assessment required prior to first-in-human clinical trials involves a composite of safety tests, each with specific objectives. The data are often combined to provide a multidimensional safety assessment, including information on exposure, and are interlinked with drug pharmacokinetics and pharmacology. Currently, data from in vivo models provide answers to key questions demanded by regulators, and some of those questions arose from past safety scares in the clinic, notably around the 3T’s: Thalidomide, Terfenadine, and TeGenero.

Of note, three key regulatory in vitro tests are currently integrated into the first-in-human safety assessment of a drug. These assess genotoxicity (in vitro Ames, Micronucleus) and QT prolongation risk (hERG assay). Tests with a single endpoint are always followed by an in vivo evaluation (in vivo micronucleus for genotoxicity and telemetry for cardiovascular) prior to multiple dose testing in humans. The guidelines do not suggest that analysis of the data from these in vitro tests would be sufficient on its own. When it comes to translation, NAMs will also face challenges similar to those in in vivo studies.  Risk assessment is not easy. Drug development is based on biology. It is non-binary with a lot of grey. There are no easy yes/no answers, and the challenge is that all decisions in the drug development process cost money. Who is ultimately going to make the decisions? One reason that nonclinical assays may not be as predictive as they should is that data is ‘looked over’. This demonstrates how challenging it will be to convert a complex biological system into a simple binary in vitro test system.

Many of the NAMs models being developed focus on one organ. They  sometimes measure only one endpoint, such as cytotoxicity, an endpoint for which translation to an in vivo situation is not so simple (Gould and Templin, 2023).  Shrimali et al (2025) showed how the better NAMs models for liver-induced drug injury (itself a complex, nonsingular target) were likely not so much related to the model per se but linked to the number of endpoints measured. The more endpoints, the more predictive. So, it’s critical to consider how tests are set up, what data they are generating, how to translate the data, and what decisions you want to draw from the data.

The hERG assay has a nice ‘cut off’ margin, but concerns have arisen that good molecules are being thrown away too early. Additional tests have since been added to address this using the add-on CiPA assays.  In vivo studies can also face challenges with translation, and it might be surprising to look more closely at how a potential issue is noted, then ignored, or written off too early and too quickly. 

There is no doubt that NAMs represent promising tools for reducing in vivo studies. Today, they already play a key role in early drug development, primarily as a screening tool to make early decisions or as a mechanistic tool. The next step will likely involve their use alongside traditional methods. For some, if the endpoint is clear, they could be validated to use as a stand-alone solution for hazard identification in a human risk assessment. As we develop and implement these tests using language-learning models, the way we do drug development will change. We have already seen how the cosmetic industry has made key changes in its use of in vitro models. While regulators are providing some guidance, it’s going to take a concerted team effort across the regulatory, pharmaceutical, and CRO sectors to bring meaningful change. Input will be required from discovery scientists, toxicologists, pharmacologists, pharmacokineticists, and regulators.

Can we do it? Absolutely. But let’s ensure the science stays ahead of the hype.

Sarah Gould, PhD, is Director and Senior Principal Scientific Advisor at Charles River Laboratories. She has more than 25 years of experience in both biotech and blue-chip international pharmaceutical companies. Sarah has delivered multiple toxicology and safety pharmacology programs and regulatory dossiers globally from Discovery to License, for small and large molecules, gene therapy, vaccines/adjuvants, with various regulatory interactions.

References:

1.    EMA (2025) Reflection paper on non-human primates in safety  testing of human medicinal products and opportunities  for 3Rs implementation
2.    FDA (March 2026)  General Considerations for the Use of 1 New Approach Methodologies in Drug Development  Guidance for Industry
3.    FDA  (March 2026)  Draft Guidance on Alternatives to Animal Testing in Drug Development. News release.
4.    MHRA (2025) Replacing animals in science: A strategy to support the development, validation and uptake of alternative methods
5.    Shrimali et al (2025).  New Approach Methodologies (NAMs) for drug induced liver injury (DILI): where are we now? Drug Discovery Today: 30 (9)