If you followed the headlines over the past two years, you could be forgiven for thinking animal testing in drug development is already on its way out. In December 2022, the FDA Modernization Act 2.0 removed the blanket legal requirement for animal studies to support an Investigational New Drug application. In April 2025, the FDA published a roadmap describing a three-to-five-year ambition to make animal studies the exception rather than the rule [1]. Commissioners have called it a paradigm shift. Investors have started asking founders whether they can skip animal work entirely.
As someone who designs the nonclinical programs regulators actually review, I want to offer a more grounded read — not because the technology isn’t promising, but because the gap between the policy signal and the day-to-day reality of getting a drug into humans is still very wide. New Approach Methodologies, or NAMs, are real, useful, and improving. They are also not about to replace animals across the board, and pretending otherwise sets founders up for expensive surprises.
What NAMs actually are
NAMs is an umbrella term for a set of non-animal methods: advanced in vitro systems, organ-on-a-chip and microphysiological systems, human-derived cell and organoid models, and computational or AI-based approaches. Their appeal is straightforward. Animal models are imperfect predictors of human biology, and the failure rate downstream is staggering — historically, more than 90 percent of drugs that clear animal studies still fail in human trials, most often on safety or efficacy grounds [6]. Methods built directly on human biology hold out the promise of catching those failures earlier and cheaper.
That promise is worth taking seriously. It is also not the same as being ready to stand alone in a regulatory submission today.
What the law actually changed — and what it didn’t
The most common misreading I encounter is that the FDA Modernization Act 2.0 banned or ended animal testing. It did not. It removed the statutory requirement that animal studies must be used, and clarified that non-animal alternatives can be used to support an IND [7]. That is a permission, not a mandate, and not a prohibition on animals. The default pathways, the data expectations, and the reviewers’ risk tolerance did not transform overnight.
The FDA’s 2025 roadmap is deliberately staged. It begins with monoclonal antibodies, where human-relevant models are furthest along, before expanding to other biologics and eventually to new chemical entities — small molecules like the ones many early-stage companies are built around [1]. If your program is a small molecule, the timeline that actually applies to you sits at the back of that queue, not the front.
Why regulatory acceptance lags the headlines
The bottleneck is not enthusiasm. It is validation. For a NAM to replace an animal study in a submission, it has to be demonstrably fit for purpose — robust, reliable, and reproducible enough to answer the specific regulatory question being asked. A model that beautifully predicts one liver-toxicity signal is not automatically qualified to clear a drug across every organ system and exposure scenario.
This points to the deeper reason animal testing won’t disappear soon, and it’s one the headlines rarely name: toxicology is the heart of animal testing, and toxicology is exactly where NAMs have the furthest to go. Animal studies exist to answer a hard, open-ended question — is this drug safe enough to give a human? — across every organ system, at multiple doses, over time, catching effects no one thought to look for. That whole-body, unanticipated-toxicity question is what a living system answers, and what a single organ-on-a-chip or in silico model, built to interrogate one pathway, still cannot. NAMs are genuinely strong at targeted questions: does this compound stress the liver, does it cross a membrane, is this specific pathway triggered. They are far weaker at the integrated safety question that defines a toxicology program. Until a NAM can replace that — not one endpoint, but the whole safety picture — animal toxicology stays. That is not a policy gap waiting on a signature. It is a biology gap waiting on science.
This is why infrastructure work matters more than any single breakthrough. The FDA, through ICCVAM, is building a shared database of validated methods to standardize what counts as acceptable. Qualification frameworks are being drafted. These are the unglamorous, essential steps that turn a promising assay into something a reviewer will accept in place of an animal, and they take years — not because anyone is dragging their feet, but because getting it wrong means missing a safety signal that reaches a human being.
It is worth being clear-eyed here rather than dismissive: the direction is real and the momentum is genuine, and the pace of guidance in 2025 and 2026 shows it. In March 2026, the FDA issued draft guidance, General Considerations for the Use of New Approach Methodologies in Drug Development, laying out four core validation principles and a framework for when NAM data can support a submission [2]. In late 2025 it drafted streamlined nonclinical guidance for certain monoclonal antibodies [3], and in May 2026 it followed with draft guidance on streamlined nonclinical safety studies for oncology biologics and conjugated products [4][5]. These are concrete steps, not just aspiration.
Two of those documents are worth flagging for anyone who works in tox, because they hint at where this is heading. Both the monoclonal antibody draft and the oncology draft state that when a product has similar pharmacological activity in both rodent and non-rodent species, general toxicology may be conducted in a single rodent species, supported by a weight-of-evidence risk assessment, rather than the traditional two-species rodent-plus-non-rodent design [3][4]. That is a meaningful reduction. But note the conditions: it is still draft, it is limited to specific product classes, it leans heavily on prior knowledge of the target, and it still requires animal data. It reduces animals; it does not eliminate them. That is the pattern to expect across the board — narrow, conditional, evidence-heavy reductions, not wholesale replacement.
The international picture complicates the timeline further
Drug development is rarely a single-country exercise. Even as the FDA moves, a sponsor planning to file in multiple regions has to satisfy other regulators whose acceptance of NAMs is on its own timeline. Harmonization through bodies like ICH and OECD is progressing, but it is slow by design, because every agency has to independently satisfy itself that a method is reliable before staking human safety on it. A NAM that is qualified for a specific purpose with one agency is not automatically portable to another. For a global program, that means the practical floor for animal data is set not by the most progressive regulator, but by the most conservative one in your filing strategy. That reality alone pushes full replacement further out than domestic headlines suggest.
What this means if you’re building a drug right now
For founders and development teams, the practical guidance is less dramatic than either the boosters or the skeptics suggest. NAMs are already valuable for internal decision-making: screening candidates, prioritizing series, de-risking a program, and filling data gaps where a traditional model is unavailable or unethical. Using them there is smart and costs you nothing in regulatory standing.
The mistake is assuming you can build your IND-enabling package around NAMs alone and expect a smooth review today. For most programs, especially small molecules, the realistic near-term model is complementary: run NAMs alongside the animal studies regulators still expect, use the human-relevant data to strengthen your story and catch problems early, and let the two datasets reinforce each other. Over time, as specific methods earn qualification for specific questions, the animal component shrinks. That transition will happen endpoint by endpoint, not all at once.
A reasonable interim approach that several observers have proposed is exactly this kind of parallel path — continuing to generate animal data while NAMs accumulate the track record needed to show they are at least as reliable for a given purpose [8]. That is not a failure of ambition. It is how you replace a safety-critical system responsibly.
The honest version of good news
The reason I push back on the “animal testing is over” framing is not that I’m attached to the old way. It’s that overpromising damages the very transition everyone claims to want. When a founder budgets and plans as if animal studies are optional today, and then meets the actual data expectations at their pre-IND meeting, the result is lost time, lost money, and a hard lesson about the difference between what the law permits and what a review division will accept.
The better story is the true one. NAMs are advancing, the regulatory system is being rebuilt to accommodate them, and the trajectory points toward genuinely less animal use over the coming decade. But “over the coming decade, method by validated method” is a very different plan than “skip it now.” Founders who understand that distinction will make sharper decisions, avoid costly detours, and still be early to adopt these tools where they actually help. That is the grounded optimism this moment calls for — not a pipe dream, and not a dismissal, but a realistic map of a real transition already underway.
Author Bio
