Animals as Models

Nell Kriesberg
Division of Multidisciplinary Studies
North Carolina State University

Introduction: the idea of a model

As noted in Section One Tutorial One, animals have been used as stand-ins for human beings since ancient times: as our investigative tools have become more refined, our ideas about models and their uses have also evolved. The idea of using a model to test our knowledge of and extend the parameters of what we know about the world is basic to the scientific method. As noted in a recent article in FASEB Journal,

Models are an indispensable manipulation of the scientific method; as deductively manipulatable constructs, they are essential to the evolution of theory from observation….the term model is used synonymously with analog to mean that which is similar in function but differs in structure and origin from that which is modeled….Thus, an analog both simplifies and puts into familiar terms a complicated phenomenon and hence enables one to think much more clearly about the subject; things are more ‘intuitively’ obvious.” (Principles and philosophy of modeling in biomedical research, Massoud, et. al., FASEB Journal, 1998, p. 275-6)

This similarity in function is critical because a model is at best a copy, an imitation or representation in some fashion of what we are examining. Utilizing a model for studying phenomena is based on two fundamental tenets of the scientific method. First is the central method of the experimental approach, that of manipulating only one variable at a time in order to establish a causal connection between two events.

Second is the idea that we can understand the whole by examining the pieces and then putting them back together, the reductionist approach. This supports one of the central ideas of using a model: that although the copy may be imperfect, as long as the specific part of the whole we are examining is analogous, this will work. If we are using a biological model, a living organism of some sort, we are furthermore basing our work on the assumption of physiological similarities between species and the idea of a unifying biochemistry as well. Thus, in using models from different species to answer questions about both our own and other species, we are assuming a high level of evolutionary conservatism .

When speaking about the choice of biological models to answer specific questions Avril D. Woodhead relates the story of the discovery of the citric acid cycle and notes,

The crucial experiments of Krebs demonstrated the catalytic effect of C4-dicarboxylic and C6-dicarboxylic acids on oxygen uptake by minced pigeon breast muscle. This preparation was chosen after Szent-Gyorgi showed its activity in the metabolism of C4-dicarboxylic acids; moreover, pigeons were plentiful in Sheffield. Kreb’s results and interpretation were questioned by Breusch because he could not find the catalytic effect of citric acid on the oxygen uptake of minced cat muscle. Subsequently, Kreb’s results were verified, and the existence of the cycle was firmly established. In retrospect, the difference between both muscle preparations is based on the fact that pigeon breast muscle is particularly rich in mitochondria, as opposed to the white cat muscle. (Arnost Kleinzeller, “Introduction: The Choice of Non-mammalian Animal Models for Biomedical Research” in Non-mammalian Animal Models for Biomedical Research, Avril D. Woodhead and Katherine Vivirito, Editors, Boca Raton, CRC Press, 1989, p. 3)

This is an example of two basic issues in using models, the conservation of biochemistry among species and physiological differences between species: this tension between what can be extrapolated and what cannot is one of the challenges of working with models. Even if one leaves out the moral calculus, to spend time and resources on a model that cannot answer the question(s) at hand, is unethical because it is a waste of someone’s time and materials.

Animal models as subset of the idea of models in general

Animal models are actually a subset of the larger idea of the model: there are in vitro models, cell lines, as well as mathematical and computer models. 

Biomedical models can be of many types-from animal models of human disease to in vitro, or animal or modeling systems for studying any aspect of human biology or disease. (Biomedical Model Definition in Biomedical Models and Resources: Current Needs and Future Opportunities, National Academies Press, 1988, p. 10)

Types of in vitro models include cell cultures (e.g. testing mitochondrial function or cell lysis,) micro-organism studies (e.g. the AMES test which uses strains of Salmonella typhimurium and rat liver microsomes) and isolated tissue assays (e.g.TESTSKIN, where human keratinocytes are put into a collagen base.)

The reason for in vitro methods is not just an ethical or humane based concern; often these non whole animal methods have their own distinct advantages such as the ability to reduce variability, allow for greater control of environment, provide for large numbers of testing strata with less expense, require less chemical materials and thus also produce less toxic waste, and can be quicker and cheaper than in vivo methods. It can also be simpler to isolate the variables with in vitro methods. Yet, if we are to address the challenge of extrapolating to a complex system, the model needs to present enough complexity to be useful. There needs to be some sort of balance between the ability to isolate the phenomenon, to single out variables (the reductionist aspect) and the fidelity to the complexity of the target system. In the article previously cited Massoud et. al. note some attributes of an ideal model, tying such a model integrally into issues of statistical design and hypothesis formulation:

Each biomedical investigation should begin with a clearly defined hypothesis. A model is devised to implement this hypothesis and allow investigation (11). All scientific modeling proceeds from the `principle of contradiction': `A and non-A' is an invalid statement (4). Beyond this point, the logical force of a model is exactly the same as that of a scientific hypothesis generally (4): to be logically forceful it must be capable of undergoing tests that may falsify it (the well-known criterion of Karl Popper; ref 13). Furthermore, the model should be: 1) heuristic in nature (a good `fit' both to the hypothesis and the available data), i.e., be appropriate to the primary features of the real world, 2) permit application of the available/desired techniques for its manipulation (vide infra), and 3) accessible to evaluation by a specified set of criteria measures (vide infra) (11). Proving a hypothesis to be untrue is not a complete defeat so long as the investigation was carried out in a scientific manner. These negative findings serve to prevent future fruitless searches in that same specific direction. Review of the work may also offer new hypotheses and even indicate new models that could prove to be successful in the future. (11) (Principles and philosophy of modeling in biomedical research, Massoud, op.cit. p. 277)

If we think of the ideal model as an instrument or tool that can both explicate and make causal connections between observed events, it is clear that there is no general, universal ideal model: what is ideal will be very specific both to the question(s) being asked and the practical situation of the researcher. For example, in genetic research, Dictyostelium discoideum, a soil living amoeba with six chromosomes may fulfill experimental needs since, “The estimated number of genes in the genome is 8,000 to 10,000 and many of the known genes show a high degree of sequence similarity to genes in vertebrate species.”(Dictyostelium discoideum, Model Organisms for Biomedical Research, NIH) In this case there may be a high level of discrimination for these particular genes and differing levels of fidelity to different target organisms, depending on the complexity of the target organism.

Key members of the research team

Using an animal as a model will necessarily involve a team approach.

It is useful to think in terms of a team approach in the choosing of an animal model and the planning of the experiment in general. The principle investigator is certainly the key responsible individual, as the scientific expert and also the supervisor of the laboratory and the research staff and students. Other key members of the team include the animal care staff, who are experienced in meeting the logistical needs of an animal-based study and can advise on potential limitations in available support. The laboratory animal veterinarian is trained in health and welfare issues associated with the use of research animals, but is also knowledgeable about animal models and comparative biology. Finally, a statistician may be an invaluable part of the team. (Rick Fish, Unit Five - Animal Models and Biomethodology)

The laboratory animal veterinarian, having an expert knowledge of different species, is the first person to go to with questions about appropriate choice of an animal model; he or she will also have an array of experience about minimizing pain and distress for all species and information about how best to undertake different procedures to minimize discomfort. For research animals, housing and environmental concerns are often as important as procedural questions and the veterinarian will have advice in this area as well. Perhaps some projects will need the advice of a specialist in comparative medicine as well. The laboratory animal veterinarian is also a required member of the IACUC.

The other key team member is the statistician. The statistician is an expert in determining early on what is necessary for the work to have sufficient statistical power to ensure rigorous results. This will be invaluable when publication time arrives, thus avoiding the criticism that many published articles do not show proper statistical analysis. To use statistics improperly is unethical, even if not intentional, since it will mislead others and ultimately, waste the experimental resources.

Starting with a clear specification of the objective of the experiment, the scientist should consider the ethical implication of the study, the validity of the model, sources of variation, and how some of these variables can be controlled using formal, appropriately randomized, experimental designs. It is important to choose carefully the independent variable(s) or treatments, the dependent variables (characters or outcomes), and the appropriate numbers of animals. The proposed method of statistical analysis should be determined at the design stage. However, the data should be studied for possible anomalies and may require transformation to a different scale before statistical analysis using parametric or nonparametric methods. Finally, the results should be presented to clarify exactly what was done and what was the outcome, "warts" and all.

The design, analysis, and interpretation of biomedical experiments are best performed with the aid of a good statistical textbook, dedicated statistical software, and advice from a statistician. Anyone reading and understanding the papers presented here cannot fail to do better, more humane animal experiments in the future. (Michael F.W. Festing, “Introduction: The Design and Statistical Analysis of Animal Experiments,” Issue on Experimental Design and Statistics in Biomedical Research, ILAR Journal, Volume 43, Number 4, 2002, p. 192)

The statistician will have expertise is such matters as the correct number of animals and how to make decisions about organizing the experimental unit in order to achieve the greatest degree of statistical rigor. For example, a cage of mice can be one unit or the individual mouse can be a unit. Further, if the skin of the mouse is shaved and several agents applied to the skin patch, then the skin patch is the unit. One can combine a number of units into a larger experimental group, or break down the units into subsets or mini-trials. Each approach is statistically different, an example of statistical refinement. Both the veterinarian and the statistician will have information about minimizing experimental variability among animals.

Classic discussions of statistical methods for analyzing results from experiments using animal subjects often do not include any mention of animal welfare or ethical concerns. The intent of the 3R’s however, is to bring these concerns into the experimental paradigm; if we consider animals as vulnerable populations, individuals at risk when used as experimental tools, than the principles of Replacement, Reduction, and Refinement become ethical imperatives. Shamoo and Resnik add two more R’s to this list, Relevance (emphasizing that research protocols need to specifically balance the harms and risks to animals with the benefit accruing to both animals and people) and Redundancy avoidance (doing a thorough literature search at the planning stage.) Others have noted a sixth R, that of Responsibility; consultation with a laboratory animal veterinarian is part of Responsible research using animals. And, as noted above, in considering a seventh R, Regulations, legally, a veterinarian must be consulted in the planning of any experiments with the potential to cause pain and/or distress.

The argument over fidelity and extrapolation

Fidelity is how close a model is to the organism or condition we are studying in our target species. For instance, chimpanzees have a high fidelity to humans. Unlike other species they can be infected with the HIV virus, but they don’t develop signs of the infection. In this case, cats, which exhibit a lower fidelity to humans than chimpanzees, are a better model for HIV in humans. Feline Immunodeficiency Virus disease, (FIV) caused by a virus from the same family as HIV, causes an illness in cats that is similar in symptoms and progression to HIV in humans. Thus, cats may exhibit a low fidelity to humans in overall anatomy but show a high degree of extrapolation as models for AIDS.

On the other hand, in a study of estrogen receptors, yeast is a good model as this particular hormonal system exhibits a high degree of evolutionary conservation. Obviously, though, this is an example of low fidelity to our species. The fact that studies on estrogen receptors in yeast can be extrapolated to human beings shows a high level of discrimination for that particular hormonal system. Issues of extrapolation are central to the choice of a model. The idea of discrimination emphasizes again the reductionist approach of the scientific method. It is clear that there are not too many features in common between yeast and humans, and yet, according to the reductionist model, as long as the piece being examined shows high discrimination, then it can be studied “as if” the piece belonged to the target organism.

One of the goals of identifying complete genomic databases for different species is to improve extrapolation. Many vertebrate species show a high degree of genetic conservation. If a particular set of symptoms is similar in two different species that also share genetic loci for that particular symptom pattern, then the scientific approach of examining variables as part of the whole makes sense. An example of this is using the mouse for cardiac studies:

The capacity to selectively mutate genes or create excessive or deleted gene expression has given researchers the possibility to evaluate the significance of certain gene product for structure-function studies of cardiac proteins and their role in heart disease. To date, several hundred mutant mouse strains and also a few mutant rat strains have been generated (http://tbase.jax.org). The number of genetically engineered mouse lines for cardiovascular research has been growing rapidly. Mouse is currently the model organism studied most using transgenic approach, since mice breed rapidly, the maintenance costs are lower, and the general knowledge of mouse genetics is at a high level. Germ line transmission has first been achieved in mouse embryonic stem cells. In larger mammals, such as rat, microinjection is the most widely used method (Mullins & Mullins 1996). (Genetically engineered animal models in cardiovascular research, Literature review, Oulu University Library, 2002)

Yet this approach has critics, both within the scientific community and without. In a recent Journal of NIH Research, Jessica Bolker and Rudolf Raff commented,

Model species are often discussed as though they represent a linear continuum of complexity and, by implication, a linear evolutionary order. The real phylogenetic tree is busy and has multiple crowns separated by varying evolutionary times…Recent experiments have led to the hopeful suggestion that new techniques for studying tissue induction in zebrafish may enable us to dispense with most of our current small group of models and instead focus all our attention on a single species, zebrafish, from which we could learn everything about vertebrate development.   We believe the opposite is true: we need to broaden our focus and study a larger variety of organisms. In order to assess the assumptions that support the model-systems paradigm, we have to go beyond it and consider comparative data from a wider range of species—in particular, species not selected according to the common set of criteria that characterizes the current models. Continuing to reduce the number of species we study makes the supposed universality of our observations an increasingly tenuous hypothesis. (Jessica A. Bolker and Rudolf A. Raff, “Beyond Worms, Flies, and Mice: It’s Time to Widen The Scope of Developmental Biology,” in The Journal of NIH Research, June 1997, volume 9, p. 36-7)

In terms of relevance to ethics, this is an interesting comment since it relates to the Russell and Burch’s concept of the 3Rs as a viable approach when using animal models. If one of the ideas behind Replacement is to “step down” to a species lower on the evolutionary scale (e.g. substituting a rat for a pig) for ethical reasons and yet maintain the ability to extrapolate, then Bolker and Raff’s concerns are important. If the research that is done on a rat cannot be completely extrapolated, then the use of the rat might be ethically problematic in terms of how sufficient the data might be; on the other hand, some of the data might be useful in terms of basic research. It is true that “no significant results” for a particular study will still both add to the general store of information and prevent another researcher from going down that path. As an example, pigs fed a high fat diet develop an atherosclerotic disease process showing symptoms similar to those in humans; rats do not. And yet there may still be reasons to explore diet related physiological changes in rats; even if just for the rats’ benefit. This complex issue of the relationship(s) between basic and applied research is beyond the scope of this website.

Bolker and Raff voice another sort of concern, one that others have raised as well, and that is the difficulty of extrapolating a complex system from a simpler one. Going back to the example of estrogen receptors in yeast, although the receptors themselves may display a high degree of discrimination, what of the exceedingly more complex system involving estrogen receptors in mammals? If the reductionism inherent in studying a process in a simpler organism cannot tell us what we need to know, this is an ethical problem because valuable resources have been wasted. As noted above, given the need for basic research, the challenge of extrapolating from a simple system to a more complex one is still an ongoing challenge for researchers. 

Hugh La Follette and Niall Shanks, both in the Philosophy Department at East Tennessee State University, argue against the use of animal models, saying that extrapolation from animal to human is questionable because of evolutionary changes between species, part of the adaptation of complex species to different environments. They emphasize the need for greater reliance on human subjects research, both clinical and epidemiological, saying that fidelity is basic to valid extrapolation.

Thus, Claude Bernard's pioneering work in physiology -- and especially his methodological prescriptions -- strongly influenced the institution of a paradigm governing biomedical research using animals. This paradigm, uncontaminated by the appearance of evolutionary theory, has guided the practice of biomedical research for most of the twentieth century. The costs of researchers continued acceptance of the Bernardian paradigm are substantial. Physiology's continued insensitivity to evolution has led it further and further away from the other biological sciences. And it has likely hindered medical advance by insisting on a single-minded methodology which assumes all significant advances come from laboratory experiments on non-human animals, and which downplays the significance of clinical and epidemiological studies. (Hugh La Follette and Niall Shanks, Animal Experimentation: the Legacy of Claude Bernard,  International Studies in the Philosophy of Science, 1994, p. 12)

The argument over the proper relationship of clinical and epidemiological studies to experimental studies is an historic one in the history of science. Bernard’s original exhortation to researchers to focus on provable causal events in the laboratory was an effort to establish specific natural laws for physiological processes as a prior step to studying treatment. We can see several strands of methodological disagreement here. There is the issue of the reductionist approach to studying basic science before turning to applications; this also involves the search for an appropriate model to study the phenomenon. And then there is the question of the validity of the overall reductionist approach as opposed to a more holistic one which also brings up the dilemma of appropriate model. There is currently a continuing debate on the complex question of extrapolating from the animal model to the human one.

If one decides that the highest fidelity organism for humans is other humans there are another set of obvious critical ethical concerns about issues of informed consent, fair balancing of risks and benefits and the rights of vulnerable populations. The Belmont Report is one of several classic documents (The Nuremberg Code and The Helsinki Declaration are two others) that attempt to set specific guidelines about the ethical constraints in using human beings as research subjects. The current scientific model for the process of biomedical research includes both animal and human subjects. Interestingly enough, the legal battle over constraints using human subjects has a more recent history.

Those who question the use of animals in research say that issues of fidelity and extrapolation are not the point. They say that such issues are not moral questions. The animal rights approach says that the issue is not whether the research can benefit humans, but rather is it is moral to use animals as instruments for our use in the first place. As Tom Regan notes,

There is only one serious moral defense of vivisection. That defense proceeds as follows. Human beings are better off because of vivisection. ...One thing should be immediately obvious. The benefits argument has absolutely no logical bearing on the debate over animal rights. Clearly, all that the benefits argument could possibly show is that vivisection on nonhuman animals benefits human beings. What this argument cannot show is that vivisecting animals for this purpose is morally justified. Whether animals have rights is not a question that can be answered by saying how much vivisection benefits human beings. (italics in original, Tom Regan, Empty Cages: Facing the Challenge of Animal Rights , New York: Rowman & Littlefield Publishers, Inc. 2004, p. 174)

The ideal animal model  

It is the very complexity of live animals that make them desirable as models. So, is the only correct model for human beings another human being? One of the difficulties with using humans in research is that our complexity may interfere with getting specific data for scientific questions. Thinking back to the example of research about the relationship between a high fat diet and atherosclerotic disease, the latter is a highly complex, multi-factorial illness. To make a specific causal connection between only a high fat diet and atherosclerosis, to get discrete information, a simpler, less confounding biological model is needed.

No dictionary describes an animal model, much less an animal model of disease, and so we can define it for our purposes here as a living organism with inherited, naturally acquired, or induced pathological process that in one or more respects closely resembles the same phenomenon occurring in man. Animal models, in this sense, never provide final answers but offer only approximations, for no single animal model can ever duplicate a disease in man. Thus, animal models should not be expected to be ideal, nor to be universally suited to all foreseeable uses. On the other hand, for a model to be a good one, it must provide a new insight, have relevance to a particular problem and respond predictably. (Stanford Wessler, M.D., Introduction: What is a Model? In Animal Models of Thrombosis and Hemorrhagic Disease, Bethesda, NIH, 1976, p. xi)

What should an ideal model provide the researcher? (We reproduce this list from NetVet from their section describing the pig as a model, but it is clear that any species, or for that matter, a cell line or a one celled organism such as the amoeba described above, could fit this description)

  1.   Accurately mimic the desired function or disease
  2.   Species availability
  3.   Data extrapolatable to man
  4.   Be available to multiple investigators
  5.   Be handled easily by most investigators
  6.   Survive long enough to be functional
  7.   Fit available animal housing facilities
  8.   Be of sufficient size to provide multiple samples
  9.   Be polytococous (multiparous) so that multiple offspring are produced for each gestation
(NetVet, Animal Models for Human Disease: Swine) 

There are a number of discussions of different animal species as “the ideal biomedical model,” but it seems clear that the goodness of fit of a specific animal model is ideal only in terms of a specific protocol. When working with a whole live animal two general aspects to consider are statistical challenges (e.g. variation and sample size) and challenges of welfare.

The challenge of variability or “Factors That May Influence Animal Research”

The reality of any complex biological system (even that of an inbred mouse) is that there will be variation among individuals; this variation can be either genetic vs. the variation associated with experimental conditions and techniques.

All animal models are subject to biological variation as a result of genetic and nongenetic variation and the interaction between them. Even genetically identical littermates will vary to some extent as a result of chance developmental effects, social hierarchy, and unequal exposure to environmental influences. Good experimental design aims to control this variation so that it does not obscure any treatment effect, with the statistical analysis being designed to extract all useful information and take into account any remaining variation. (Festing, The Design and Statistical Analysis of Animal Experiments, ibid.p. 191)

Different variables can be utilized more fully as part of an overall design involving blocking, nesting, and so forth. This increases the overall benefit of the model chosen, since 1) there is a balance of using the animal to obtain more rather than less information; 2) the research will have a higher degree of rigor, another ethical “good;” and 3) the open-endedness of basic research interests can be addressed, along with the specific goals of the project. 

The challenge is to separate out the level of variation inherent to a live animal—a range desired since this mimics the natural biological system in a natural world, the reason for using a whole animal in the first place—from that variation that will result in confounded research results. In Chapter 29, “Factors That May Influence Animal Research,” (Laboratory Animal Medicine, 2nd Edition, Fox, et. al.) authors Neil S. Lipman and Scott E. Perkins note,

…the multitude of complicating factors that have been described in the literature. The reader should understand that there are likely additional factors yet to be recognized, as well as interactions among factors that may also influence experimental outcomes…In order to obtain reliable, meaningful results, an attempt should be made to control or standardize all known biological, environmental, and social factors when conducting experiments involving animals.(Lipman and Perkins, op.cit. p.1143)

Lipman and Perkins divide these factors into “intrinsic” and “extrinsic considerations.” Intrinsic considerations include items such as genetics, age, sex, immune and nutritional status, circadian rhythms and endocrine factors.  Extrinsic factors include all the physical components of housing (temperature, humidity and air exchanges, noise, light, radiation, vibration and caging and housing sorts of issues.)  Lipman and Perkins also includein this category items such as diet and water, as well as any use of pharmaceuticals. Actual biological pathogens such as viruses, bacteria, parasites and fungi are also considered external chemical factors in their listing, as well as pheromones, noting that this is a sexual-social chemically based form of normal communication. A recent publication from The National Academies Press addresses this challenge of variation in terms of mice and rats, particularly genetically engineered rodents:

My definition of "biological integrity" is incomplete, but, for the moment, consider the term to mean "the stability of intrinsic and extrinsic factors that define the structural and functional characteristics of an animal." Therefore, the benchmarks for defining a laboratory rodent in the era of genetic engineering must include at least the establishment, standardization, and monitoring of factors such as genotype, phenotype, microbial status, and environmental quality. Criteria such as reproductive capacity and other health-related factors such as susceptibility to infection should also be considered.

These concepts also imply that biological integrity can be perturbed by intrinsic or extrinsic interference, which may be overt or subtle. This threat is especially relevant considering the diversity of settings in which genetically engineered rodents are being made. Variability can be caused by genetic drift; the influence of genetic background on the penetrance of a phenotypic trait; opportunistic infection that may be pathogenic, disruptive to normal responses, or conducive to erroneous phenotyping; environmental stresses such as noise, vibration, and threatening odors; and many other factors. Variability also can be abetted by diverse or ill-defined terminology. For example, and as noted elsewhere in these proceedings (Lindsey, 1999), the term "specific pathogen free" has lost value because of the lack of precision with which it often is employed and perceived. Additionally, the increased use of animals inherently increases risks to biological integrity from dense housing and increasing exchanges of animals and animal products among laboratories, nationally and internationally. (The National Academies Press, Microbial and Phenotypic Definition of Rats and Mice: Proceedings of the 1998 US/Japan Conference, 1999, ILAR)

Further, both the veterinarian and the statistician need to be consulted in planning any sort of refinement to a study. To take just one example, implementing a program of environmental enrichment will necessarily add to the complexity of extrinsic factors. Mice have strong hierarchical social patterns and adding complexity to their environment may result in larger variation of responses due to differing sorts of interactions within the cage population. This variation will impact the statistics of the project. In an example of “research on research,” a recent report notes,

Environmental enrichment in the standardized cages of laboratory animals has a beneficial effect on animal welfare. However, it has been argued that enrichment may increase the intra-group variability in experimental results, and thus increase the number of animals necessary to produce statistically significant data. The Research Foundation is pleased to present this successful project (Nr. 66-99), which was designed to provide solid experimental evidence to help settle this controversy. (Environmental enrichment does not affect the variability of animal experimentation data in the Light/Dark Test, 3R Research Foundation)

The veterinarian will have specific suggestions for environmental enrichment for different species to reduce this variability between individuals and the statistician will have the expertise to factor the variability into the overall statistical plan for the project.

An example of a detail of variation that might escape notice is one of cage placement within an animal room. A sample case study from The Biomedical Investigator’s Handbook: for Researchers Using Animal Models (Foundation for Biomedical Research, 1987) describes a situation where previously healthy mice began dying as a metabolism experiment proceeded. A consulting veterinarian, a rodent specialist, was consulted and reported that:

“A few days of monitoring temperature, humidity, ventilation rate, and ventilation pattern inside the environmental chamber revealed the problem. Although warm air rises, a peculiarity of the ventilation system in the chambers resulted in higher air temperatures in the mouse cages occupying the bottom rows of the shelves….Exchanges of air in the lower cages were not frequent enough to remove the heat generated by mice. The veterinarian concluded that the mice were dying from heat stress. The air distribution system within the environmental chamber was modified and the use of the bottom shelves was limited. The study resumed with no further problems. (from Chapter 2, “Unwanted Variables: Preventing Complicating Factors,” op. cit. p. 15)

Another example of variation would be the amount of human interaction: if the animals are getting specific treatments involving handling this becomes critical. A number of studies have shown that repeated interactions with trained staff can lower the fear threshold; this is particularly important if the protocol calls for dosing, injections, and other forms of physical manipulation. As noted in the Journal of Applied Animal Welfare Science,

The promotion of affection towards laboratory animals has scientific and empirical underpinning. It has been shown in rabbits that frequent, gentle handling lessens the animals' fear response during stressful situations (Anderson, Denenberg, & Zarrow, 1972; Kertsen, Meijsser, & Metz, 1989). Rabbits who receive special positive attention from personnel show a markedly increased resistance to the development of atherosclerosis compared to subjects who receive no extra attention (Nerem, Levensque, & Cornhill, 1980). Regular gentle handling has a protective effect on the experimental induction of stomach ulcers in rats (Weininger, 1954). (Viktor Reinhardt, “Compassion for Laboratory Animals: Impairment or Refinement of Research Methodology” Journal of Applied Animal Welfare Science, 2003, (6)2, p. 128.)

There is another discussion of variability, put forth by Kenneth Shapiro. He says that the goal of reducing individual variation as part of the overal scientific approach is at heart, suspect. The very nature of a living animal is the reality of individual variety.

Consider a rat that is chronically implanted with an electrode in his or her brain and is connected by a tether to machinery that sends stimuli and receives and records responses. The rat is more a part of the instrumentation than a discrete object of study. The animal is merely a conduit for certain energies, fluids, and electrical impulses....The animal is not viewed or acted toward as a whole creature let alone as an individual....The chronically implanted animal is not so much put "under the microscope" as made part of it....The attempt to enhance observation through instrumentation, to record results n strictly quantitative terms, and to remove the person and bias of the investigator, all in the service of achieving objective and positive (certain) results, raises the question: What is the relation between model and modeled? When we selectively breed, genetically engineer, deindividuate, despecify, and deanimalize an animal, do we know what we have left? (Kenneth J. Shapiro, "A Rodent For Your Thoughts: The Social Construction of Animal Models," in Animals in Human Histories: The Mirror of Nature and Culture, Mary J. Henninger-Voss, editor, Rochester: University of Rochester Press, 2002, p. 455)

Shapiro's critique goes deeper than a concern with variation. He, like Regan, is questioning the morality of using an animal as an instrument.

Searching for animal models

Choosing animals as models is a complex process:

Basic laboratory animal science is concerned with the quality of animals as sentient tools in biomedical research…. The choice or selection of animal model depends on a number of factors relating to the hypothesis to be tested but also on more practical aspects associated with the project and with the project staff and experimental facilities. The usefulness of a laboratory animal model should be judged on how well it answers the specific questions it is being used to answer, rather than on how well it mimics the human disease. (Jann Hau, “Animal Model”, Chapter 1, in Handbook of Laboratory Animal Science, 2nd Edition, Volume II, Jann Hau and Gerald Van Hooser, Editors, Boca Raton, CRC Press, 2003, p. 1 and p. 7-8)

In thinking about animal models of disease, there are two kinds of models: 1) those where the disease or condition occurs spontaneously in the animal due to genetic programming and 2) an induced model where we have created the disease or condition either via genetic manipulation or direct injection or surgical manipulation. 

Considerations in Choosing an Animal Model:
Adequate discrimination (data extrapolatable to target species)
• Adequate fidelity (necessary anatomic structures, biochemical pathways, etc.)
• Good literature base; historical usage (i.e., accepted animal model)
• Readily available to other researchers
• Genetic and microbiological characterization
• Sufficient size for obtaining necessary samples (blood, urine, biopsies, etc.) and substance administration
• Accommodation in existing animal facilities (caging, environmental controls, exercise and environmental enrichments)
• Experience/training of animal care staff
• Experience/training of research staff
• Tractable; trainable
• Low cost (purchase; maintenance)
• Good reproductive performance
• Minimal indigenous disease (or ability to control)
• Ethical considerations
• Public relations implications
(Rick Fish, Humane Care and Use of Animals in Research - Unit Five - Models and Biomethodology)

The Institute for Laboratory Animal Resources has a search engine for animal models.

The National Institutes of Health has a website focusing on Model Organisms for Biomedical Research

Balancing harms and benefits: monitoring pain, stress and distress

As Richard Fish notes in The Humane Care and Use of Animals in Research: Unit Three - Pain and Distress:

Many procedures performed in research animals can rightly be assumed to be painful, based on their ability to cause pain in humans; this is the language in U.S. Government Principle IV. However, there is also legitimate need to customize pain and distress treatment in individual animals, and this requires tools to recognize signs of pain and distress.

A frequently-cited paper (Morton DB, Griffiths PHM. 1985. Guidelines on the recognition of pain and discomfort in experimental animals and an hypothesis for assessment. Vet Rec 116:431-436) proposed a set of observations for assessing pain and distress; these included change in body weight, external physical appearance, clinical signs, and changes in behavior. A refinement of this approach has been described by E. Carstens and Gary P. Moberg in an article (Recognizing Pain and Distress in Laboratory Animals) in a recent ILAR Journal focused on ” Humane Endpoints for Animals Used in Biomedical Research and Testing. Carstens and Moberg suggest evaluation of pain in three categories: general behavior (e.g., activity, appetite), appearance (e.g., self-grooming, posture), and physiology (e.g., body temperature, respiratory pattern).

The assessment of distress presents additional difficulty. While there are physiological changes characteristic of distress that could be measured, they often involve obtaining samples (e.g., blood) by methods that may themselves introduce stress. Like pain, there are behavioral correlates of stress, but they are not well characterized; importantly, there is even less known about how to distinguish stress, which is a normal part of life, from distress.

The assessment of pain and distress in animals requires knowledge of both normal behavior and those behaviors that might indicate pain or distress. Those who are responsible for routine animal care, as well as research staff who are involved with potentially painful or distressful experimentation, should seek training to help with this important task. An excellent starting point is Chapter 4 (Recognition and Assessment of Pain, Stress and Distress) in the Recognition and Alleviation of Pain and Distress in Laboratory Animals.

One of the areas for Refinement is the establishing of humane end points that protect the welfare of the animals, preventing undue suffering if at all possible. This is of particular importance in protocols where pain and discomfort become part of the study, e.g. in some behavioral studies, toxicology, wound healing, pain research, etc. One of the problems with using a live animal model for invasive protocols is that it brings into stark relief the effort to balance clear risks with appropriate benefits. Here is where the exhortation to reduce the number of animals used may meet with another sort of ethical conundrum. Is it better to cause more pain to a few as opposed to less pain to many?

There is philosophical debate about whether it is better to cause more suffering to fewer animals or less suffering for many to achieve the scientific end. That situation is not common in practice, but the U.K. law takes the view that the level of individual suffering is what matters and thus harms should always be minimized. (David B. Morton, “The importance of non-statistical design in refining animal experiments,”  in Applied Ethics in Animal Research: Philosophy, Regulation and Laboratory Applications, John P. Gluck et. al, Editors, West Lafayette, Purdue University Press, 2002, p. 174-175)

Morton notes that score sheets are a means to increase the “rigor” of monitoring pain and distress during a protocol. The technicians responsible for the daily care thus become an integral part of the study team, reporting early stages of difficulty so that interventions can be sooner rather than later. This practice actually improves the welfare of the staff as well, since the people responsible for the animals on a daily basis are now empowered. Items for scoring include behavioral details (for mice in this case) such as “walking on tiptoe,” “hunching,” “grooming” “not inquisitive and alert” and “vocalization when palpated,” or “amount of jelly mash eaten.” Morton notes that the score sheet will be unique to the study and the model species used, another example of seeing the research animals as individuals. Kansas State University has posted Pain and Distress Observation Worksheets; also see their Post Op Animal Health Score Sheet

The question of "pain, stress and suffering" of animal subjects necessarily brings up again, the overall ethical issues of the cost and benefits of animal research. As Regan has noted, whether or not the particular use of an animal model benefits humans is a separate question from the overall morality of using the animal as a tool in the first place. In discussing the use of animal models in psychology research, Kenneth Shapiro, executive director of the Psychologists for the Ethical Treatment of Animals, describes the possibility of using both the rights and utilitarian approaches in attempting to decide on specific protocols.

One philosopher combines utilitarian and rights principles. In response to current political realities, Rollin suggests a two tier ethic in which first a utilitarian and then a rights principle are applied, in that order. If the benefits of a proposed study are judged to "clearly outweigh" the costs, the rights principle is then invoked to prohibit any animal suffering beyond that intrinsic to the experimental procedure. Of course, the last qualifier is significant as it means that the experimental procedure is preemptive; the rights of the animals are safeguarded only outside the demands of the experiment. The discussion to this point suggests that, in practice, this would not be much of a gain for the animals.

However, alternatively, the two principles could be combined giving priority to rights. Elsewhere, I have suggested that any proposed research that employs an experimental procedure that is "intrinsically objectionable" be prohibited. One rating scale measuring degree of invasiveness prohibits procedures that fall into the most severe category. British regulations also embody this notion of prohibiting certain procedures. This concept supersedes utilitarian considerations by precluding certain procedures, in principle, independent of any possible resulting benefits. In effect, this concept trumps the rights of an individual against the possible benefits to the group. Only after this rights-based criterion is met is the proposed research then judged on utilitarian grounds, along the extensive lines I have described. This mixed ethic is more true to the spirit of a deontological or rights philosophy in that it makes a certain principle preemptive. At the same time, it takes advantage of the pragmatic power of utilitarianism, the weighing of practical results--arguably, its most attractive feature. (Kenneth Joel Shapiro, Animal Models of Human Psychology: Critique of Science, Ethics and Policy, Seattle: Hogrefe & Huber Publishers, 1998, p. 282

Study Questions

1. In thinking about Kenneth Shapiro's description of a "two tiered" ethical approach, which makes more sense to you, thinking first of the utilitarian principle and then a rights one, or vice versa? In answering this question, first consider an oncomouse and then consider an "oncochimp." Do you find this "tiered" approach helpful or not? Why or why not? See "Patenting Animals: The Harvard Oncomouse" in The Human Use of Animals: Case Studies in Ethical Choice, F. Barbara Orlans, et. al., (New York: Oxford University Press, 1998) for ideas.

2. It is standard practice to do a literature search before begininng to plan a new experimental protocol with an animal model. Quite often, the fact that there is a history of using a particular animal as a model, with a collection of published articles already in existence, gives support to the choice of that animal model. Give some ethical and scientific reasons in support of following this tradition; then, give some reasons against following this tradition in a new protocol.

Discussion Question

Creating an animal model of chronic disease brings up difficult ethical questions: Genetically modified animal models may be a special case for ethical concern, particularly in issues of husbandry, monitoring and humane endpoints.. "Oncomice" are mice that are genetically engineered to develop cancerous disease at an early age. Recalling our earlier discussion about the team approach in planning and care of animal models, what major concerns and areas of emphasis would personnel such as the principle investigator, the statistician, the laboratory animal veterinarian, and the animal care staff all have to offer in developing a monitoring plan? In thinking over your answer, refer to The National Academies Press book Animal Biotechnology: Science Based Concerns (2002) that has two chapters that discuss these situations: Animal Health and Welfare and Animals Engineered for Human Health Purposes. You might also want to consult Recognition and Alleviation of Pain and Distress in Laboratory Animals or the score sheet links noted in the "Balancing harms and benefits" section of this tutorial.

Additional resources

Richard Fish, Humane Care and Use of Animals in Research: Unit Five - Animal Models and Biomethodology

Institute for Laboratory Animal Research (ILAR) Journal Online has many articles available concerning models

Basic texts on animal models:

Jann Hau and Gerald Van Hooser, Editors Handbook of Laboratory Animal Science, 2nd Edition, (Boca Raton, CRC Press, 2003 )  See especially: Jann Hau, “Animal Models, Chapter 1, Volume I;  Michael F.W. Festing and Benjamin J. Weigler, Experimental Design and Statistical Analysis,” Chapter 14 and David B. Morton and Jann Hau, “Welfare Assessment and Humane Endpoints,” Chapter 18, Volume II.

The National Academies Press, Biomedical Models and Resources: Current Needs and Future Opportunities, ILAR, 1998

Fred W. Quimby, “Animal Models in Biomedical Research,” Chapter 30 in Laboratory Animal Medicine, 2nd Edition (NY: Elsevier Science, 2002)

W.M.S. Russell and R.W.Burch, The Principles of Humane Experimental Technique

Websites:

American Heart Association: Circulation Research: Animal Models of Human Disease

American Federation for Aging Research: Animal Models Information Center: and  “What animal models of aging are used in research?”

National Cancer Institute: Animal Models Initiative

National Institute of Health: Mammalian Gene Collection

Institute of Human Virology: Animal Models

Zetetica: Why Genomics May Be a Bad Model For Biological Research