Regular readers of this blog will know that the governance of the UK Government’s Replacing Animals in Science strategy has been weighing heavily on my mind.

How will we know whether it is working? What does meaningful progress look like? What should the Key Performance Indicators actually measure?

These things matter because any strategy needs more than a direction of travel. It needs a destination. Governance should keep us on the path; KPIs should tell us whether we are getting closer. And in this case the Government has, on the face of it, given us a remarkably clear destination: the elimination of the use of animals in science “in all but exceptional circumstances”.

The difficulty is that those final three words are doing an enormous amount of work.

What, exactly, is an exceptional circumstance?

At present, the strategy gives us only a limited indication, in the context of those tests identified as having potential for rapid transition to non-animal methods. There, the Government acknowledges that animal tests may still be required where other countries do not accept alternative approaches and animal data are needed to satisfy global regulatory requirements.

That may explain one particular circumstance in which animal use could continue during a transition. It does not give us a definition capable of supporting the strategy’s overall ambition.

And without one, there is a real danger that “exceptional” becomes so elastic that almost everything we currently do can be fitted inside it.

Starting from a different question

The first principle, I think, is that exceptional cannot simply mean “where no validated alternative exists”.

That sounds superficially reasonable, but it reproduces one of the fundamental weaknesses of the current system: the assumption that an animal experiment remains the default until somebody produces a formally validated like for like replacement, that replicates the outcomes of the animal ‘model’.

Increasingly, that is not how good science works.

A non-animal approach may consist of several different sources of evidence — human cells and tissues, organoids, computational modelling, mechanistic information, epidemiology, existing human data and other methods — brought together to answer a particular question. What matters is not whether we can replace an animal experiment with another test that produces the same outcome. What matters is whether we can answer the scientific or regulatory question within the context of use to the level of confidence actually required.

There is an equally important second half to this argument. A shortcoming in a non-animal method does not, by itself, demonstrate the value of an animal experiment.

Animal models also have limitations. If the proposed animal experiment is poorly predictive of the human response, or unlikely to resolve the uncertainty that actually matters, then the fact that the non-animal evidence is imperfect cannot magically make the animal evidence good.

So I think we need to flip the question.

Rather than asking, “Is there an alternative to this animal experiment?”, we should ask:

What decision are we trying to make, what evidence do we need to make it, and is animal use genuinely necessary to provide that evidence?

A proposed definition

As a starting point for discussion, I would suggest that an exceptional circumstance exists only where, for a precisely defined and socially permissible scientific or regulatory purpose, an independent assessment concludes that:

1. The decision-relevant question cannot presently be answered to the required standard of confidence by any reasonably available non-animal approach, or combination of approaches.
2. The unresolved uncertainty is material to preventing a serious threat to human health, animal health or the environment, or to meeting another compelling public-interest need.
3. The proposed animal procedure is itself scientifically capable of materially reducing that uncertainty, and is more decision-relevant than the available non-animal evidence.
4. The expected public benefit is sufficiently important and sufficiently probable to justify the animal harms involved.
5. The need has not arisen because of avoidable delay, lack of investment, institutional custom, commercial convenience, duplication or publication pressure.
6. Any authorisation is limited in both scope and duration, subject to review, and accompanied by a plan to remove the need for the same exception in future.

The crucial point is that these conditions must be cumulative. It should not be enough to satisfy one or two of them. An application should have to pass every test.

Nor should we ever designate an entire area of science as “exceptional”. Cancer research is not an exceptional circumstance. Neuroscience is not an exceptional circumstance. Basic research is not an exceptional circumstance. A specific unresolved question, attached to a sufficiently important purpose and incapable of being answered adequately in another way, might be.

That distinction is essential.

Four practical gates

Six principles are useful, but if this is to mean anything in practice it also needs to be usable by regulators, Animal Welfare and Ethical Review Bodies and researchers themselves.

I think the test can therefore be reduced to four sequential gates.

Gate 1: Is the purpose genuinely important enough?

Before considering methods, we should ask whether the purpose itself could justify exceptional animal use.

An urgent and serious public-health threat might. Scientific curiosity alone should not.

That is not an argument against curiosity-driven science. It is an argument that once animal use becomes exceptional rather than routine, not every otherwise legitimate scientific purpose will necessarily cross the threshold required to justify it.

Gate 2: Can the question be answered without new animal use?

This should require consideration of the full available evidence strategy, as well as the existing knowledge base.

What relevant data already exist? What human-relevant methods could be used? Could several approaches be combined? Could data or samples be shared? Could the question itself be reframed?

If the answer can be obtained to the required standard without animals, the application goes no further.

Gate 3: Will the animal experiment materially improve the decision?

This is the gate I think is most often missing from current thinking.

“No adequate NAM exists” does not automatically mean “therefore use an animal”.

The applicant should also have to demonstrate that the proposed animal experiment has a realistic prospect of resolving the uncertainty that matters. If the model is poorly predictive, the result would be ambiguous, or the outcome would not change what happens next, then the experiment should not be authorised simply because we have historically been comfortable doing it.

Sometimes the scientifically honest answer may be that neither the animal nor non-animal method can answer the question well enough; a difficult but important outcome to recognise in a resource-scarce environment.

Gate 4: Is the exception proportionate and temporary?

Finally, an exception should never become a permanent category of activity.

If it is authorised, it should be tightly drawn, time-limited and reviewed. Repeated applications relying on the same evidence gap should themselves trigger investment, method development or regulatory action to close that gap.

Because otherwise it is no longer exceptional.

Flipping the legal presumption

None of this necessarily requires us to throw away the framework created by the Animals (Scientific Procedures) Act. In many respects, it builds upon principles already there, but it does require us to change the starting assumption. At present we ask whether a proposed animal procedure is justified based on the harm-benefit trade-off, and whether alternatives have been adequately considered.

An “all but exceptional circumstances” system would require asking something different:

Animal use is presumed not to be available. What precisely makes this case exceptional enough to justify departure from that presumption?

That is a much higher bar.

And if “all but exceptional circumstances” really is the destination of the Government’s strategy, it has to be.