Articles

The regulatory door just opened for new approach methodologies

FDA’s draft guidance on non-animal methods runs on two principles: context of use and weight of evidence. Together they tell you exactly where a predictive model is allowed to sit.

2 min read

It is funny how this new FDA draft guidance connects directly with the virtual cell model I was writing about the same morning.

FDA released draft guidance on general considerations for the use of new approach methodologies in drug development, or NAMs, covering in vitro, in silico and other non-animal methods. The framework they lay out is exactly the lens I had been applying.

Two principles run through the whole document

Context of use. A fit-for-purpose NAM, even if not fully validated, may adequately address specific toxicological concerns, but it is always considered alongside everything else you know about the product.

Weight of evidence. Nothing stands alone. The method is weighed with the rest of the package.

That is precisely where a predictive model fits right now. Useful for mechanism. Useful for early risk thinking. Useful for asking better questions before you spend a lot of money. FDA is actively encouraging sponsors to use NAMs in regulatory submissions, especially where they improve predictivity and human relevance of nonclinical testing.

But not something you rely on by itself.

Vioela’s note: Context of use is not a phrase to skim past. It is the single sentence that defines what question the method answers, for which kind of product, at which stage. A method accepted for one context is not accepted in general, and reviewers will hold you to the exact words. Write that sentence before you generate any data, and bring it to the agency early, at a pre-IND meeting, while changing course is still cheap. The expensive version of this lesson is a finished study that answers a question slightly different from the one the reviewer needed.

The validation point

The guidance is also clear that for NAM data to be useful for regulatory decision-making, the data must be reliable and there must be confidence in the methodology, confidence that is typically built through validation.

Which brings me back to the point I keep making about predictive models: retrospective validation would go a long way toward establishing that confidence.

Vioela’s note: Some background on why this door opened now. Until the end of 2022, the law itself required animal testing before human trials. Congress removed that requirement, and in 2025 FDA published a roadmap for reducing animal testing, starting with antibody drugs. The shift is real, but the practical path for most sponsors is still side by side: run the new method alongside the conventional study and let the comparison build the confidence. Replacing an animal study outright, on a new method’s word alone, is still the exception.

The actual question

So in a way this is less about whether the model works, and more about whether it is being used in the right place, for the right question, with appropriate supporting evidence around it.

Thinking tool, not decision tool. At least for now.

But the regulatory door is open. That part is new.

Adapted from a post first published on LinkedIn.