A Hypothesis-Based Approach: The Use of Animals in Mental Health Research
This Director’s Message on the use of animal models in mental health research is one of two focused on this topic and is aimed at the research community. A companion message, written for a general audience, is titled: What Can Animals Tell Us About Mental Illnesses?
Model organisms play a crucial role in modern neuroscience. Exploring the function of molecules, cells, circuits, and systems and how they relate to behavior often requires the use of methods to examine the intact brain that, for ethical and practical reasons, can only be performed in animals. Yet, the translation of results gained from animals into treatments for humans has been challenging. At the National Institute of Mental Health (NIMH), we have been engaged in a year-long interactive process to understand and clarify the appropriate role of animal research in our portfolio. While we continue to refine our policies, this message and an accompanying notice are meant to communicate our current policies regarding the scientifically justified use of animals in mental health research.
Utility rather than validity: A hypothesis-based approach
First and foremost, we must realize that there is no such thing as a true animal model of a psychiatric disorder. Such models are, at best, approximations. How then should we evaluate animal models? For the past few years, we at NIMH have been asking the following questions about studies involving model organisms:
- What is the question being asked?
- Is it an important question?
- Does the proposed experimental system enable that question to be answered?
These questions support an evaluation of the appropriateness and impact of a hypothesis-based approach to studying the biology of mental illnesses using animal models.
What is the question being asked?
Researchers create models to test hypotheses. We believe that NIMH-supported studies should not be aimed at establishing whether a particular model has validity for understanding a human mental illness. Instead, they should be aimed at testing a specific hypothesis. Too frequently, we receive proposals in which an investigator proposes to create an animal model of a disorder and “validate” that animal model by running a series of assays to document similarities between the animal and humans with the disorder. Such studies are problematic for many reasons, not the least of which is low statistical rigor, given that positive publication bias coupled with multiple tests can often lead investigators (and those that follow) astray. A strong focus on a mechanistic hypothesis, with the animal model designed to test that hypothesis, coupled with a rigorously planned and sufficiently powered experimental design, increases the reproducibility of the results. Moreover, focusing on the question ensures that the knowledge gained can be built upon by succeeding investigators.
Is it an important question?
For the purposes of grant applications, we urge that investigators explicitly set forth the hypothesis upfront and explain how the proposed experiments address this hypothesis. Notably, the hypothesis can be driven by basic science interests. How does disrupting a given cellular function alter the development of a neural circuit? What is the role of that circuit element in a particular behavior? How do multiple brain regions interact during that behavior? They may also be driven by clinical questions, testing hypotheses that arise from studies in patients. Do disruptions in inhibitory interneurons alter prefrontal function? What is the role of sensory inputs to the amygdala in social recognition of emotion? How does exposure to adverse environments during development affect cortico-limbic interactions? Each of these questions addresses fundamental areas of biology of relevance to mental illnesses.
There are some areas of neurobiology where we have identified specific priority areas and additional topics where we might require detailed justification given known challenges. Several of these areas are delineated in the accompanying notice. For example, fundamental studies of the biology of genes and gene products implicated in mental illnesses by unbiased approaches using genome-wide significance thresholds are of particular interest to NIMH given their definitive relationship to the human conditions. Conversely, genes previously identified via candidate approaches and not subsequently verified by genome-wide approaches are of considerably less interest. These and other priorities in the area of genetics and genomics are clearly stated in guidance we have published following the Report of the National Advisory Mental Health Council Genomics Workgroup.
Similarly, we are prioritizing computational behavioral phenotypes over simplistic, pharmacologically-validated behavioral tests for reasons to do with specificity and clarity of mechanism. For example, traditional behavioral responses to stress paradigms are particularly problematic. Non-specific tests such as the forced-swim or tail suspension tests, among others, have largely failed to reveal translatable neural mechanisms, and lack specificity from a pharmacologic-validity perspective. Approaches that examine and rigorously quantify the impacts of stress on reward and arousal systems, by contrast, are promising avenues of research that hold considerable promise to reveal novel mechanistic insights and lead to new therapeutic avenues.
Does the experimental system proposed enable that question to be answered?
The issue of the usefulness of an animal model cannot be divorced from the question that the model is trying to answer. Precision here is crucial, as are the measured variables. For example, consider facial expressions of emotion. Mice don’t have them. But mice do have measurable facial expressions of physical pain. If an investigator wants to understand the neural pathways leading to facial expressions of emotion, mouse models will not help. But if the goal is to understand the pathways leading from pain to facial motor output and how they are modulated during behavior, the mouse might be useful.
The point here is that one should ask the question first and then figure out what the best experimental system is to answer that question. The evaluation of the experimental system should include consideration of the ethical and efficient use of resources, the feasibility of the approach, and the potential evolutionary conservation of the mechanism of interest. Most questions will probably require a variety of systems in order to answer them fully, maximize rigor and reproducibility, and ensure translatability. For many investigators, this will mean collaborations and team science. But for investigators using animals, this means clarifying why the particular model organism was chosen and how that model facilitates hypothesis testing.
We are here to help
We have attempted to make the NIMH position on animal models as clear as possible, but there are a number of gray areas, and every investigator’s situation is unique. In addition to examining the resources we have available on the web — including this message, the notice, and other materials — I encourage as always communication with NIMH program staff early in the grant preparation process to ensure that potential issues are addressed appropriately in advance.