The wall between veterinary and human oncology is not a biological fact. It is an industrial accident — and it is costing us time, data, and lives.
Built by history. Not by biology.
Two industries grew up separately. One organized around human patients, human trials, human regulatory pathways. The other around animal health — livestock economics, companion animal welfare, veterinary training. Different journals, different grant bodies, different career tracks that almost never intersected.
Over time, the separation hardened into something that felt like principle: the idea that what happens in an animal body is categorically less relevant to human medicine. This was never a scientific conclusion. No study established it. It was a historical artifact dressed in the borrowed authority of long habit.
The biology does not know about the wall. The wall is on our side.
Mammals share the fundamental architecture of cancer. Tumor microenvironments operate according to conserved mechanisms. Immune evasion, cellular dysregulation, the cascade of failures that allows malignancy to take hold — these are not uniquely human phenomena. The molecular vocabulary of cancer is largely the same across species.
The dominant model for preclinical oncology has been the engineered mouse: cells implanted into an immunocompromised animal under controlled laboratory conditions. These models are useful. They are also profoundly artificial. The tumor grows in an organism whose immune system has been deliberately disabled to prevent rejection. The environment that generates the data bears no resemblance to the environment a therapy will eventually enter.
The result is a field with an extraordinary rate of preclinical success and a troubling rate of clinical failure — therapies that perform brilliantly in forced models, then encounter immune systems that are working and tumor microenvironments that are real.
Immunocompromised host. Implanted tumor. Controlled, artificial conditions. Produces clean data that tells us less than we assume — because the environment is nothing like the one the therapy must eventually enter.
Intact immune system. Real tumor microenvironment. Biology that mirrors human pathology. A therapy validated here has earned its data in a way no forced model can replicate. The conditions were real.
Companion animals develop cancers spontaneously — not implanted, not induced, not engineered. Real disease arising naturally in real immune systems, in animals that live in our homes and age on timescales we can observe.
The most important principle of the OneHealth methodology is this: the animal is treated because it is sick and treatment is the right thing to do. Translational value is not the reason for the care. It is what we learn along the way.
This is what OneHealth means in practice: not a metaphor about interconnected ecosystems, but a concrete methodology in which companion animal oncology generates human-relevant data as a natural consequence of doing right by the animal.
The science and the conscience of the work agree on the same course of action. When that happens — when ethics and epistemology point in the same direction — it is usually a sign the approach is sound.
Companion animals first. Not because the regulatory pathway is shorter — it demands the same rigor as human trials. But because the science is better: the data carries a biological credibility that the conventional preclinical sequence cannot provide.
In practice this means running companion animal and human development as parallel programmes — not sequentially, but simultaneously. The animal programme provides translational signal that informs and accelerates the human track. One programme running across species, generating regulatory-ready data as a natural output.
Those who called the companion animal pathway a detour were not wrong that it was different. They were wrong that different meant slower. A therapy proven in an animal with a real, spontaneous tumor enters human translation carrying a level of confidence that no forced model provides. The longer route is, by the standard that matters, the faster one.
Remove the wall — not by ignoring regulatory realities, but by building a programme that runs deliberately across both sides — and the field of the possible changes. Translation accelerates because the preclinical data is more predictive. Costs fall because one programme generates signal across two tracks simultaneously. The animals are treated as patients rather than instruments.
The wall costs us time, data, and the ability to recognize when a finding in one species reveals something true about another. What we lose when it comes down is only the wall.
Part of the BioFund Concept Paper series. ← One Problem · Foundations · Two Clocks →