If you are designing preclinical work right now, the ground under it has moved more in the last eighteen months than in the previous twenty years. Most of the coverage has been either breathless or dismissive, so here is what actually happened, followed by the part that affects what you do next.
The sequence of events
The legal groundwork came first. The FDA Modernization Act 2.0, passed in late 2022, explicitly authorised non-animal alternatives to support an investigational new drug application, and removed the animal-study requirement for biosimilar applications. Importantly, it authorised alternatives rather than mandating them, which is why little changed immediately.
What changed things was the FDA roadmap to reducing animal testing in preclinical safety studies, published in April 2025. It set out a stepwise transition with timelines and validation frameworks, and it began deliberately with monoclonal antibodies, where animal models have long been a poor predictor of human response.
Then it accelerated. A December 2025 draft guidance addressed reducing non-human primate testing for certain monoclonal antibodies. On 18 March 2026 the FDA released comprehensive draft guidance on New Approach Methodologies. Alongside it, NIH announced $150 million in Complement-ARIE awards and a NAMs Data Hub. The FDA published a one-year progress report in April 2026.
The legislation has not caught up. The FDA Modernization Act 3.0 passed the Senate in December 2025 and the House in July 2026, but as separate bills, so it still needs reconciling before it can be signed. Anyone telling you animal testing has been banned is wrong on both the law and the guidance.
What counts as a NAM
The term is broader than most people assume. It covers complex 2D and 3D in vitro systems including organoids and organs-on-chips, in silico simulation, chemical reactivity studies, and work in phylogenetically lower organisms such as zebrafish.
That breadth matters, because it means the shift is not simply "organoids replace mice". In practice the accepted arguments are combinations: a computational model plus a human-relevant in vitro system plus existing clinical data, assessed together as a weight of evidence.
The detail almost everyone gets wrong
The March 2026 draft guidance from CDER states that neither validation nor formal qualification is required for a NAM to be used in support of a drug application.
This surprises people, and it cuts both ways. It removes the circular trap where nothing could be used until it was validated and nothing could be validated until it was used. But it does not mean any assay will be accepted. It shifts the burden onto the sponsor to justify fitness for purpose for that specific question. There is no approved list to work from, which is harder in practice than a list would be.
What this means if you are in an academic lab
Be clear that this is regulatory guidance for drug development, not a rule about academic research. Your IACUC protocol is unaffected. Nobody is going to stop you running mice.
The second-order effects are the real ones:
- Method-development work has a clearer destination. A human-relevant model built to answer a safety or efficacy question now has a defined regulatory context, which is a stronger framing for a grant or a paper than it had two years ago.
- Characterisation is the bottleneck, not novelty. A weight-of-evidence argument needs models with documented reproducibility and known limits. Careful benchmarking against human data has become more valuable relative to building yet another novel system.
- The skills are shifting. Organoid culture, microphysiological systems, and quantitative in vitro to in vivo extrapolation are more employable than they were, in both directions between academia and industry.
- Funding is following. The NIH Complement-ARIE programme exists specifically to develop and validate these methods.
Where organoids genuinely substitute, and where they do not
Being honest about this is what separates a useful model from an overclaimed one.
They do well where the biology is tissue-intrinsic and human specificity matters: barrier function, target engagement in human cells, many forms of organ-specific toxicity, and disease modelling in patient-derived material where no good animal model exists.
They do poorly where the answer depends on a whole organism. Anything needing circulation, an intact immune system, or true pharmacokinetics across organs is not currently replaceable by a dish, and the guidance does not pretend otherwise. Most organoids also lack vasculature and immune components, which is precisely where a lot of interesting toxicity lives.
What to actually do
If you are planning in vivo work now, the useful question is not "should I avoid animals?" It is which specific question in your study is best answered by which system, and whether some of your planned animal work is answering something a human-relevant model would answer better.
That is a stronger position for a thesis committee than either extreme, and it happens to be the position the guidance is actually asking for.