Research Article: A standardised framework to identify optimal animal models for efficacy assessment in drug development

Date Published: June 13, 2019

Publisher: Public Library of Science

Author(s): Guilherme S. Ferreira, Désirée H. Veening-Griffioen, Wouter P. C. Boon, Ellen H. M. Moors, Christine C. Gispen-de Wied, Huub Schellekens, Peter J. K. van Meer, Malcolm R. Macleod.


Poor translation of efficacy data derived from animal models can lead to clinical trials unlikely to benefit patients–or even put them at risk–and is a potential contributor to costly and unnecessary attrition in drug development.

To develop a tool to assess, validate and compare the clinical translatability of animal models used for the preliminary assessment of efficacy.

We performed a scoping review to identify the key aspects used to validate animal models. Eight domains (Epidemiology, Symptomatology and Natural History–SNH, Genetic, Biochemistry, Aetiology, Histology, Pharmacology and Endpoints) were identified. We drafted questions to evaluate the different facets of human disease simulation. We designed the Framework to Identify Models of Disease (FIMD) to include standardised instructions, a weighting and scoring system to compare models as well as factors to help interpret model similarity and evidence uncertainty. We also added a reporting quality and risk of bias assessment of drug intervention studies in the Pharmacological Validation domain. A web-based survey was conducted with experts from different stakeholders to gather input on the framework. We conducted a pilot study of the validation in two models for Type 2 Diabetes (T2D)–the ZDF rat and db/db mouse. Finally, we present a full validation and comparison of two animal models for Duchenne Muscular Dystrophy (DMD): the mdx mouse and GRMD dog. We show that there are significant differences between the mdx mouse and the GRMD dog, the latter mimicking the human epidemiological, SNH, and histological aspects to a greater extent than the mouse despite the overall lack of published data.

FIMD facilitates drug development by serving as the basis to select the most relevant model that can provide meaningful data and is more likely to generate translatable results to progress drug candidates to the clinic.

Partial Text

The use of non-human animals to evaluate the safety of new drugs is an integral part of the regulatory research and development process [1,2]. Established at a time when laboratory animals were one of the most complex systems available, they are still considered as the gold standard. However, despite their apparent value as a drug testing system to predict safety and efficacy in humans, scientists are increasingly aware of their considerable drawbacks and limited predictivity [3–6].

The existing approaches for assessing external validity cannot be used by researchers to find what is the most relevant model to demonstrate the efficacy of a drug based on its mechanism of action and indication. Here, we present a method to assess the various aspects of the external validity of efficacy models in an integrated manner–the Framework to Identify Models of Disease (FIMD). A ‘model of disease’ is here used for any animal model that simulates a human condition for which a drug can be developed.

In FIMD, the validation of an animal model relies on the definition of the disease parameters and therefore, it depends on how well-understood the aetiology and pathophysiology are. Poorly understood diseases (e.g. Alzheimer’s) represent a major challenge for the definition of these parameters and thus to the validation of animal models in these indications. However, FIMD provides researchers with a platform to discuss both the more established aspects of a disease as well as the ones in which there is no consensus in a structured and standardised manner (e.g. aetiology, genetic, biochemical). This way, an animal model’s ability to simulate the different aspects of disease according to various theories or hypotheses (whenever available) can be discussed and assessed in one place by the same parameters.

Based on a drug’s mechanism of action, models can be first discriminated by assessing whether the correlation between non-human animal and human drug studies of relevant pathways is available in the Pharmacological Validation domain. The other domains provide additional information, such as the presence of relevant genes and biomarkers. Finally, the validation level is an index of the reliability of a model’s overall ability to mimic the human condition, serving as another layer to further differentiate potentially more useful from less useful models. The combination of all these features allows researchers to select, among a plethora of models, the model most likely to correctly predict the efficacy of a drug in humans.