New Approach Methodologies, or NAMs, are no longer a niche topic in drug development. They are increasingly part of how teams think about human relevance, translational risk, and where conventional animal studies may fall short.
NAMs include a broad range of non-animal and reduced-animal approaches: human cell-based assays, organoids, organ-on-chip systems, computational models, physiologically based pharmacokinetic modeling, biomarkers, omics, and other tools that can help interpret safety risk earlier and more mechanistically.
Used well, NAMs can answer questions that animal studies often cannot answer cleanly. They can help a team understand whether a toxicity signal is likely to be human relevant, whether one molecule in a series is safer than another, whether a specific mechanism such as mitochondrial toxicity or transporter inhibition is emerging, or whether additional work is needed before a candidate moves forward.
But NAMs are not magic. They are not a shortcut around good drug development. And for most programs today, they do not replace the GLP toxicology package.
The real value of NAMs is more practical: they help teams make better decisions earlier.
Why the Conversation Is Happening Now
The push toward alternatives to animal testing is not new. It is rooted in the 3Rs principle: replacement, reduction, and refinement of animal use. For years, NAMs were often treated as interesting technologies to monitor, pilot, or use selectively, but they were rarely central to how most drug programs were run.
That is changing.
The FDA Modernization Act 2.0 removed the long-standing statutory requirement that all new drugs be tested in animals before entering human trials. FDA has since released a roadmap describing how the agency intends to expand the use of NAMs, beginning with selected areas such as monoclonal antibodies. NIH has also made major investments in human-based research methods through programs such as Complement-ARIE.
These changes matter, but they should not be overinterpreted. The near-term message is not that animal studies are disappearing from drug development. For most small-molecule and biologic programs, animal toxicology remains a core part of the nonclinical safety package. The more immediate shift is that sponsors are increasingly expected to think carefully about where animal data are informative, where they are limited, and where human-relevant methods can add decision-relevant evidence.
The practical question for pharma and biotech companies is not “Can NAMs replace animal studies?” but “Which NAMs are worth using for this molecule, at this stage, to answer this specific decision?”
The Limits of Animal Studies Are Real
The industry has known for a long time that animal studies do not predict human outcomes as reliably as drug developers would like. Many drug candidates that look promising in preclinical testing still fail in the clinic, for reasons that include lack of efficacy, safety findings, tolerability, exposure limitations, and commercial factors.
Animal toxicology remains essential in the current regulatory framework, but it is not a complete predictor of human risk. One of the most frequently cited cross-company analyses, conducted by Olson and colleagues, showed that combined rodent and non-rodent toxicology captured many, but not all, human target-organ toxicities. A meaningful fraction of human toxicities still occurred without clear concordant animal findings.
That is the translational gap NAMs are trying to help address.
Some organ systems are particularly difficult. The liver remains one of the most persistent blind spots in drug development. Drug-induced liver injury continues to be a major reason for clinical holds, late-stage attrition, black box warnings, and market withdrawal. It is also biologically complex: mitochondrial dysfunction, bile acid transporter inhibition, reactive metabolite formation, oxidative stress, immune activation, ER stress, steatosis, phospholipidosis, and patient susceptibility can all contribute.
No single animal model captures that complexity reliably. No single NAM does either.
That is why NAMs should be viewed as evidence-generating tools, not as simple replacements.
What NAMs Actually Cover
NAMs are broader than many people realize. The conversation often focuses on organoids and organ-on-chip systems, but the practical toolkit is much wider.
Computational models include QSAR, structural alerts, machine-learning models, and endpoint-specific tools for liabilities such as genotoxicity, hERG risk, transporter inhibition, metabolic stability, and other safety-relevant properties.
Cell-based assays include primary human cells, immortalized cell lines, co-culture systems, and high-content imaging platforms used to identify cytotoxicity, mitochondrial toxicity, lipid accumulation, cholestatic stress, lysosomal changes, ER stress, and other cellular phenotypes.
Microphysiological systems include organoids and organ-on-chip platforms designed to reproduce selected aspects of human tissue structure, function, flow, or multicellular interaction.
PBPK and quantitative systems pharmacology models help connect in vitro findings to exposure, dose, tissue distribution, and species translation.
Biomarkers and omics can provide additional biological context, especially when a team is trying to understand mechanism, species relevance, or whether a finding is likely to translate to humans.
The important point is not which technology sounds most advanced. The important point is whether the method answers the question in front of the team.
A mitochondrial liability question, a BSEP or cholestasis question, a reactive metabolite question, a species-relevance question, and a patient-susceptibility question do not require the same NAM package. The wrong approach is to ask, “Should we use NAMs?” The better question is, “What decision are we trying to make, and what evidence would actually change that decision?”
Where NAMs Add the Most Value
For many companies, especially smaller biotechs, the greatest value of NAMs is in discovery and lead optimization.
That is where the data can still change the molecule that moves forward.
If a team is choosing among analogs, backups, or lead candidates, well-selected NAMs can help identify which molecule has the cleaner mechanistic profile. They can reveal whether one compound has a mitochondrial signal, whether another disrupts bile canaliculi, whether a third produces lipid accumulation or ER stress, or whether an apparent cytotoxicity signal is driven by exposure, solubility, nonspecific stress, or a more specific biological liability.
This matters because the cheapest toxicity problem is the one you avoid before candidate nomination.
By IND-enabling development, NAMs can still be valuable, but the nature of the value changes. At that stage, they are less likely to redirect the chemistry and more likely to strengthen the interpretation of the nonclinical package. They can support species selection, help explain cross-species differences, provide human-relevance evidence, inform clinical monitoring, and make the overall safety rationale more defensible.
That is important, but it is different from claiming that NAMs will eliminate required animal studies.
For most programs today, they will not.
What NAMs Cannot Do Yet
NAMs also have limitations, and those limitations need to be stated clearly.
Most NAMs model selected pieces of human biology. They may capture one organ, one cell type, one pathway, one exposure window, or one mechanism. They usually do not capture the full systemic biology of a living human. Immune, endocrine, nervous system, metabolic, microbiome, and multi-organ interactions remain difficult to model at scale.
Exposure is another major limitation. A positive signal in a human cell system is not automatically a clinical liability. A negative signal is not automatically reassuring. The result has to be interpreted in the context of concentration, duration, protein binding, metabolic competence, tissue exposure, projected human Cmax or AUC, and the intended clinical use.
This is where many NAM strategies fail. They generate data, but the data are not linked to exposure or to a decision.
Technical maturity also varies widely. Some assays are well-characterized and reproducible. Others are still evolving. Organoids and organ-on-chip systems can be biologically impressive, but they may be expensive, low-throughput, variable, and difficult to validate across sites. Computational models can be powerful, but they depend heavily on the quality of the training data, endpoint definitions, chemical space, and context of use.
Interpretation is often the biggest gap. Producing NAM data is not the hard part. Understanding what the data mean for a real drug program is the hard part.
How Regulators View NAMs
Regulators are interested in NAMs, but they are moving carefully.
The current regulatory posture is not “NAMs replace animals.” It is closer to this: NAMs can be persuasive when the context of use is clear, the biology is relevant, the method is technically characterized, and the result helps interpret risk in the overall nonclinical package.
That distinction matters.
A NAM result is more useful when the sponsor can explain why the model was selected, what biological question it addresses, how the assay performs, what its limitations are, and how the result changes the interpretation of risk. Regulators do not need a technology story. They need a fit-for-purpose scientific rationale.
For selected contexts, especially where animal models are poorly predictive or where human biology is clearly needed, NAMs may eventually reduce or replace certain animal studies. Monoclonal antibodies are one area where this discussion is already more advanced. But for most drug development programs today, NAMs are best positioned as part of an integrated evidence package.
They can support species selection. They can explain why an animal finding may or may not be human relevant. They can strengthen clinical monitoring plans. They can help justify why a particular risk is understood and manageable.
That is already valuable.
What Large Pharma Is Doing
Large pharmaceutical companies have been investing in NAMs for years. Many have partnered with organ-chip companies, built internal high-content imaging and omics platforms, incorporated computational safety models, and developed increasingly sophisticated translational toxicology workflows.
They are not doing this because animal studies disappear tomorrow. They are doing it because the companies that learn how to generate, interpret, and defend human-relevant evidence now will be better positioned as regulatory expectations evolve.
Large pharma also has a different use case from biotech. A large company can apply NAMs across portfolios, build internal reference datasets, benchmark platforms, and gradually integrate methods into governance. They can test many compounds, compare readouts to historical outcomes, and determine where each method is most useful.
Most biotechs cannot do that.
And they do not need to.
What This Means for Biotech Programs
For a biotech, I would not start with a platform-wide NAM strategy. I would start with the molecule, the chemistry, the target, the competitor landscape, and the next decision the team has to make.
A company with two or three molecules does not need an enterprise NAM strategy. It needs a program-specific translational safety strategy.
The questions are practical:
- What is the key safety concern for this program?
- Is it target-related, chemistry-related, modality-related, species-specific, exposure-driven, or unknown?
- Are there known liabilities in the competitor landscape?
- Are there structural alerts or mechanistic concerns in the series?
- What would make us choose one molecule over another?
- What would make the IND package more defensible?
- What would we want to know before spending money on GLP toxicology?
That is where NAMs become useful.
For example, if the program has liver-risk concerns, a human hepatocyte-based assay, mitochondrial assessment, transporter evaluation, reactive metabolite work, or high-content imaging may help identify whether the risk is real and whether it differentiates across compounds.
If the concern is cardiac liability, the right package may include hERG, ion-channel profiling, iPSC-derived cardiomyocytes, contractility, electrophysiology, or exposure-aware modeling.
If the issue is species relevance, the NAM strategy may focus on human versus animal target expression, pathway biology, biomarker translation, or cross-species pharmacology.
The right NAM is the one that informs the decision. Not the one that sounds newest.
The Investor and Partnering Angle
There is also a strategic value to using NAMs well.
For a biotech preparing for financing, partnering, or acquisition, defensible human-relevant data can strengthen the story. It shows that the team understands where animal data are informative, where they are limited, and how translational safety risk has been addressed before the program becomes expensive.
That does not mean adding a glossy organ-chip experiment to impress investors. It means generating data that a pharma diligence team would consider meaningful.
A strong NAM package can help answer questions such as:
- Why was this candidate selected over the backup?
- Is the therapeutic index supported by human-relevant biology?
- Are known class liabilities understood?
- Are the animal species appropriate?
- Are there mechanistic safety concerns that need clinical monitoring?
- If a signal appears in vivo, do we already have a framework to interpret it?
That is a much stronger story than simply saying the standard package was run.
Principles for Using NAMs Well
Start in lead optimization. That is where NAMs can most directly affect which molecule moves forward. By the time a candidate is already in IND-enabling toxicology, much of the opportunity to change the chemistry is gone.
Define the question before selecting the method. A NAM is only useful if it is tied to a decision. Do not start with the platform. Start with the problem.
Use NAMs to strengthen the evidence package, not to create a shortcut. For most programs today, GLP toxicology studies remain expected. NAMs make the package more interpretable and defensible.
Interpret results in exposure context. Concentration, protein binding, metabolism, duration, Cmax, AUC, and clinical dose matter. Without exposure, NAM results are easy to overinterpret.
Know the limitations of the model. Every NAM has boundaries. A good strategy states those boundaries clearly rather than hiding them.
Invest in interpretation. Data generation is not enough. The value comes from understanding what the result means for the program.
Engage regulators early when the NAM will be part of the submission rationale. A pre-IND discussion is often the right place to test whether the proposed approach is persuasive.
Decide molecule by molecule. Especially for biotech, the right NAM strategy depends on the target, modality, chemistry, indication, competitor landscape, and the next development decision.
Looking Ahead
The regulatory environment is moving toward greater use of NAMs, but the field needs to stay disciplined. The goal should not be to replace animal studies as quickly as possible with under-validated alternatives. The goal should be to build better, more human-relevant, more decision-ready safety packages.
Over time, the most exciting opportunity may be precision toxicology: testing compounds in models that better represent the biology of the patients who will actually receive them. That could mean disease-relevant systems, genetically diverse donor models, susceptibility models, or human datasets that help identify when a safety margin holds for the intended population and when it does not.
That is where NAMs could ultimately change the field.
But today, the message for biotech is simpler.
Use NAMs early. Use them carefully. Tie them to a specific decision. Interpret them in context. And do not oversell what they can do.
NAMs are not a shortcut around drug development.
Used well, they are a way to make drug development smarter.