Disease models in nonclinical studies

A disease model is an animal in which a condition resembling a human disease has been induced or is present genetically, so that a candidate can be tested against something other than a healthy animal. What the model reproduces is a set of measurable features of the disease, not the disease itself. When a sponsor is choosing where to place efficacy work, whether the laboratory runs the specific model the programme needs is usually the deciding factor, because establishing a model that a laboratory has not run before adds months and introduces variability that is hard to separate from a treatment effect.

What a model result establishes

An efficacy study in a disease model establishes that the candidate changed a measured feature of that model, at a stated exposure, relative to a control group. That is the whole of what it establishes. It is enough to support a decision about whether to continue, which is why these studies are run before the regulated safety package.^[S3] It is not a measurement of efficacy in patients, and a result in a model does not predict a clinical outcome.

Animal models and human cell-based systems

Systems built from human induced pluripotent stem cells address one specific limitation of animal models: the species difference in the target and in the pathways around it. Because the cells are human, a compound is tested against human protein sequences, which an animal model cannot provide. What such a system cannot provide is the whole organism, so it does not replace the animal study and it does not answer questions about distribution, metabolism or systemic effects. This article makes no claim that these systems reduce the rate at which programmes are stopped; no source establishing that could be located.

Choosing a model

The model has to contain the mechanism the drug is meant to act on, which is a narrower requirement than sharing the disease name. A compound intended to reverse fibrosis needs a model in which fibrosis is established, not one in which it is still developing. A compound acting on an immune pathway needs a model in which that pathway drives the pathology. Selection conventions exist by disease area, such as strains that develop hypertension spontaneously for cardiovascular work, and they are a starting point rather than an answer. Programmes typically begin in rodents and move to rabbits, dogs, non-human primates, pigs or goats where the question requires a larger animal, most often for anatomical reasons or for the volume of sample needed.

Four things are worth establishing with the laboratory before the model is agreed. What the historical control data look like for that model at that site, since the variability in the model determines the effect size the study can detect. How the model is confirmed to have taken, and what happens to animals in which it has not. At what point after induction the readout is taken, and why that timepoint. And how many times the laboratory has run it, since a model established for the first time on a sponsor study carries the setup variability as well as the treatment question.

Humanized animals sit between the two. In one approach human cells or tissue are engrafted into an immunodeficient animal and allowed to establish; in another, a segment of the animal genome is replaced with the human sequence. Both narrow a specific species difference rather than removing species differences generally, and both introduce their own constraints, since an animal that accepts a human graft is by definition immunodeficient. The question to ask of any humanized model is which human component it actually contains, because that determines what the result can be read to mean.