Lost in translation: Are animal models predictive of a treatment effect?

In order to satisfy the sufficiency requirement of the European Patent Convention (EPC), it is necessary for the applicant of a European patent application directed to a medical use to show that the claimed therapeutic effect is credible. However, when an patent application is filed, human clinical trial data are often not available. The claimed effect must therefore be demonstrated by data from in vitro experiments and animal models that are representative of the human disease. This practice reflects that of academic science, in which potential treatments are first tested and published using animal models before they are translated to the clinic. However, an applicant can face an uphill battle in convincing the European Patent Office (EPO) examiner that animal model data credibly demonstrate the claimed effect. 

This Kat explores the reasoning behind the use of animal models, the acknowledged limitations of animal data for predicting treatment effects in humans and the approach taken by the EPO in assessing whether such data can be used to demonstrate a medical use invention. 

Why do we use animal models of human disease? 

Model mouse

The use of non-human animals to research the pathology and treatment of human disease has obvious advantages. Testing a drug on a non-human animal is cheaper, easier and safer than experimenting on humans. Animal models of human disease can also provide valuable information about a disease or treatment at the level of the whole organism, as opposed to the artificial setting of isolated cells in a tissue culture flask. Furthermore, animal models can also be used to model isolated aspects of a complex, multi-factorial human disease, so that the underlying mechanisms of action can be determined. 

Problems with animal models of human disease

It has long been appreciated in academia that animal models of disease have considerable limitations. Animal models may, for example, be misleading with regards to how the condition operates in humans.  Particularly, the simplicity of animal models, whilst being one of the features that makes them so attractive, can also be a disadvantage. Over-simplification of the condition, for example, can lead to the identification of apparent cures that fail to translate into human patients.

Mice are one of the most widely used animal models. Much of our understanding of the human immune system comes from mouse data. However, the mouse and human immune system, whilst sharing many characteristics, diverge in many others. Some of these differences can lead to different responses in mice and human to pathogens and drugs. Indeed, more than 80% of potential therapeutics demonstrated as safe and effective in mice, reportedly fail when tested in humans (Perrin, Nature, 2014).

Modelling human disease in mice is made more difficult by the fact that, in the majority of cases, mice do not naturally suffer from the equivalent of the human disease. The disease must therefore be artificially produced in the mouse by disrupting or introducing physiological factors known to be involved in the human condition, with the hope that this will reproduce symptoms of the human disease, with varying degrees of success. There is also a growing realization that environmental conditions can be crucial in determining an organism's response to disease. This can lead to contradictory results from different laboratories using the same strain of mouse, further exacerbating the problem of translating research from mice to other organisms. 

Is translation from mouse to human obvious?

Despite the difficulties associated with translating results obtained using a mouse model into the clinic, the EPO has found that a claimed medical use is obvious in view of prior art mouse data in a number of cases.  In T 0149/93 the issue in question was whether experiments carried out in mice made the use of a particular composition to treat a human skin condition obvious. The technical board of appeal (TBA) ruled that whilst there were essential differences between the mouse and human skin, mouse experiments would have “provided a basis for possible future experiments with human beings", the problem and solution being the same.

By contrast, in T 0986/02, the TBA held that a claim directed to the use of a composition to treat systemic lupus erythematosus (SLE) was non-obvious in view of mouse data in the prior art. The TBA observed that whilst several of the prior art documents dealt with data obtained from a mouse model of the disease, these data were not consistent. A related composition to the claimed composition was seen to be protective in some mice strains whilst exacerbating disease in other strains (r. 12). The TBA thus found that it would not be obvious to a skilled person to use the claimed composition to treat SLE. The TBA further concluded that “ the mouse model for SLE is not representative for human SLE”, and that therefore data using a mouse model was not required to demonstrate the claimed effect (r. 16-18). A combination of human and in vitro data was determined to be sufficient. 

In the more recent decision T 0332/17, the question was whether a treatment demonstrated as effective in a laboratory mouse model would make it obvious to a skilled person to apply the treatment “in pets” (including pet mice). The proprietor had argued that the laboratory mice used in the prior art "did not qualify as pets, since they were albino, immunodeficient inbred strains, prone to cancer development and showing high levels of anxiety" (r 2.3.2). However the TBA agreed with the opponent's argument that "all disease models, regardless of whether they are for human or other animals [...] are intended to apply on naturally occurring diseases" (r.2.5.2), and it would therefore be obvious to a skilled person to consider the laboratory data relevant to pet mice. The TBA also dismissed an auxiliary request limiting the claims to the treatment of cats and dogs. The TBA again agreed with the opponent that, given the prior art had already made the leap from the mouse model to human, a skilled person would also expect the treatment to be applicable to cats and dogs (r. 4.2).

Comparing human and mouse data – an uphill struggle?

Where an improved therapeutic effect is the purported “problem to be solved” by a claimed invention, the EPO may reject an application even when data is provided showing a claimed therapeutic effect, because the data do not show an improved therapeutic effect over the prior art. Given that mouse models can often be said not to be directly comparable to human disease (or even to other mouse models of the same disease), this can create difficulties for an applicant using a different disease model to that of the prior art.

In T 113/11, for example, the TBA assessed the inventiveness of a claim directed to a new type of antibody to treat a symptom of SLE. The TBA noted in r. 12 and 13 that the data provided in the specification showing the effect of the antibody in a mouse model of SLE, could not be directly compared to the data provided in the closest prior art. This was because the closest prior art related to the use of a different type of antibody to treat the same symptom of SLE in a different mouse model. The TBA noted the differences in the mouse models, and indicated that the appellant had not substantiated that the data in the patent and “the model of [the prior art document] are directly interchangeable and that therefore their results are directly comparable”. As such, the data provided in the patent could not be said to demonstrate that the new type of antibody was better than the antibody of the prior art. This was despite the acceptance by the TBA that the mice used in the patent were “a known and accepted mouse model that best manifests the pathology of human SLE” (r. 5).

This type of objection is frequently seen during Examination. The EPO Examiner will argue that the data provided in support of the invention does not demonstrate that the problem to be solved in view of the prior art has been solved, because the data cannot be directly compared to that of the prior art. This Kat has seen objections from EPO Examiners in which even human data are dismissed, because the prior art uses mouse data that the Examiner argues is not directly comparable with the mouse data of the prior art. The Examiner will conclude that the improved effect of the claimed composition is therefore not demonstrated with respect to the closest prior art. However, given that an applicant may not have been previously aware of the prior art cited against them during prosecution (and what particular mouse model the prior art may use), it can be difficult to preempt such an objection by planning the appropriate experimental comparisons.

Demonstrating a surprising 
technical effect with mouse data

The applicant of a patent in the life sciences field is therefore often required to a) demonstrate why data from a mouse model is applicable to the human disease, b) simultaneously demonstrate an improved clinical effect in view of data from a disparate prior art mouse model study, whilst c) also not falling into the trap of arguing that the claimed effect would be obvious in view of these previous studies. 

As a final thought, this Kat notes that it can be difficult to objectively assess the applicability of a mouse model to the human disease it is supposed to represent. Academic researchers have an incentive to proclaim the applicability of the particular model with which they work or have developed. There is also a tendency for most research to be carried out in the model that is easiest to use, as opposed to the one that is most representative of a disease. The EPO's criteria that a mouse model should be a well-accepted model of a disease is therefore not necessarily equivalent to the mouse model being the one most representative of that disease.

By Rose Hughes