Abstract
The use of animals in biomedical and other research presents an ethical dilemma: we do not want to lose scientific benefits, nor do we want to cause laboratory animals to suffer. Scientists often refer to the potential human benefits of animal models to justify their use. However, even if this is accepted, it still needs to be argued that the same benefits could not have been achieved with a mitigated impact on animal welfare. Reducing the adverse effects of scientific protocols (‘refinement’) is therefore crucial in animal-based research. It is especially important that researchers share knowledge on how to avoid causing unnecessary suffering. We have previously demonstrated that even in studies in which animal use leads to spontaneous death, scientists often fail to report measures to minimize animal distress (Olsson et al. 2007). In this paper, we present the full results of a case study examining reports, published in peer-reviewed journals between 2003 and 2004, of experiments employing animal models to study the neurodegenerative disorder Huntington's disease. In 51 references, experiments in which animals were expected to develop motor deficits so severe that they would have difficulty eating and drinking normally were conducted, yet only three references were made to housing adaptation to facilitate food and water intake. Experiments including end-stages of the disease were reported in 14 papers, yet of these only six referred to the euthanasia of moribund animals. If the reference in scientific publications reflects the actual application of refinement, researchers do not follow the 3Rs (replacement, reduction, refinement) principle. While in some cases, it is clear that less-than-optimal techniques were used, we recognize that scientists may apply refinement without referring to it; however, if they do not include such information in publications, it suggests they find it less relevant. Journal publishing policy could play an important role: first, in ensuring that referees seriously consider whether submitted studies were indeed carried out with the smallest achievable negative impact on the animals and, secondly, in encouraging scientists to share refinements through the inclusion of a 3Rs section in papers publishing the results of animal-based research.
Animal-based research presents an ethical dilemma: if it is pursued, animals may be caused to suffer; if it is not, important biomedical benefits may be lost (e.g. Olsson et al. 2002). Scientists often refer to the scientific value and biomedical importance of animal models to justify their use (e.g. Paul & Paul 2000). However, except from a purely anthropocentric perspective (in which the effects of the experiments on the animals are ignored), ethical concern about compromised animal welfare cannot be addressed by pointing to human benefits alone. It is equally important to reduce the animal welfare problems associated with research as far as possible (see also Nuffield Council on Bioethics 2005 for a further discussion of the ethical aspects raised by the use of animals in research).
In this context, it can be useful to refer to the 3Rs (replacement, reduction, refinement) and in particular the ‘Principle of Refinement’. This principle – the third R of the 3Rs – refers to ‘any decrease in the incidence or severity of inhumane procedures applied to those animals which still have to be used’ (Russell & Burch 1959/1992, Smaje et al. 1998). To control pain with anaesthesia and analgesia may be the most obvious response when a scientist is asked to reduce animal suffering in research (although, as Richardson and Flecknell [2005] demonstrated, even in this area there is still room for considerable improvement), but other forms of suffering besides pain may also need consideration. In many neuroscience studies, such as when using animals as models for neurodegenerative diseases, pain is probably of secondary importance. Within this field, refinement should focus on limiting the negative impact of declining sensorimotor function as the disease progresses: housing adaptations (to facilitate food and water intake) and humane endpoints (which register earlier, less severe clinical signs as endpoint parameters rather than awaiting spontaneous death) are two possible approaches (e.g. Morton & Hau 2002). In the present study, we asked to what extent adverse effects of scientific protocols and refinements to address them were reported in publications reporting the use of animal models of the neurodegenerative disorder Huntington's disease (HD).
In humans, HD involves loss of striatal neurons, resulting in progressive motor, cognitive and psychiatric dysfunction. The disease is fatal, death usually being caused by disease-related complications. No cure for HD is in prospect, so treatment currently focuses on methods of reducing its symptoms (Walker 2007). In HD research, animal models (primarily rats and mice) are used to study striatal damage (which is either induced through administration of a neurotoxic substance or develops as a result of genetic intervention) and loss of sensorimotor function in varying degrees.
Materials and methods
Our analysis relates to research reports published in international peer-reviewed scientific journals during 2003–2004. A Web of Science search for ‘Huntington's disease’ for this period gave 591 hits. Of these, 92 were found to be original research reports, using animal models of the disease and available through the library services of the authors' home institutions. The selection we analysed included both in vivo studies and ex vivo studies where tissue was obtained from animals in which the clinical disease was present, but not ex vivo studies of tissue from healthy animals. Models where more than five experiments were reported were selected for a more extensive analysis of the adverse effects on the animals, such as type of induction protocol and refinement applied. The selected models were based on the administration of neurotoxic substances, quinolinic acid (hereafter QA) (Beal et al. 1986, Brouillet et al. 1999) and 3-nitropropionic (hereafter 3-NP) (Brouillet et al. 1993, 1999) to rats and mice, or on the use of genetically-modified animals, R6/1, R6/2 and N171-82Q transgenic mice (Mangiarini et al. 1996, Schilling et al. 1999, Hockly et al. 2003).
As the description of clinical signs was found to be very limited in research reports of most animal models, a number of researchers working with these models were approached personally through email – first, with a simple questionnaire, and, if necessary, with more specific questions later. Papers reporting several models were included once per model, resulting in 110 references in total and 90 references for the five selected models. Research reports using these models were analysed applying scales developed for this purpose addressing the development of clinical signs and the impact of the induction protocol, respectively (Table 1).
Severity of animal experiments
QA: quinolinic acid, 3-NP: 3-nitropropionic, BW: body weight
The scale was developed on the basis of information on clinical status obtained from own experience (R6/1 [Naver et al. 2003]), literature (R6/2 [Hockly et al. 2003], N171-82Q [Schilling et al. 1999], 3-NP [Fernagut et al. 2002] and personal communication with researcher Emerich D [QA]).
Results
Animal-based research into serious disease does not always involve severe animal distress (34 references found in categories 1 and 2; Table 2). However, the majority of experiments were found to be in the two most severe categories (36 in 3 and 17 in 4; Table 2), in which animals have sensorimotor deficits that interfere with their capacity to eat and drink normally from the cage top. Very few papers reporting such severe experiments included information on refinement measures (Table 3). Out of 14 survival studies of transgenic animals with a progressive neurodegenerative phenotype, only six referred to the application of humane endpoints and only two referred to housing adaptations (Table 3).
Distribution of models over different degrees of severity
3-NP: 3-nitropropionic, QA: quinolinic acid
Severity of clinical status expressed as number of experiments reported for each degree. Induction stress reported as median (minimum–maximum). Definitions of clinical status and induction stress are given in Table 1. Three R6/2 references were excluded because it was not possible to estimate severity from the information given in the paper
Reported refinement measures in moderate to severe experiments
NA: not applicable, QA: quinolinic acid, 3-NP: 3-nitropropionic
Data expressed as the number of experiments reporting the information in question, with the total number of experiments per model and severity indicated in the left column
In assessing experimental impact, we separated the distress caused by induction and the clinical status as regards disease development. The chemically-induced models scored highest on our scale as regards induction stress: from 2 for the implantation of subcutaneous minipumps for 3-NP administration in rats, through 3 for stereotactic injection of QA under general anaesthesia in rats, to 4 for repeated (up to 26 injections with 12 h intervals) intraperitoneal injections of 3-NP in mice. As regards clinical status, studies in which animals were kept until spontaneous death (or the euthanasia of moribund animals) scored highest on our scale. These included 3-NP administration in mice with mortality (both unpredicted in protocols that were altered accordingly [Saydoff et al. 2003, Diguet et al. 2004] and planned lethality protocols [Blum et al. 2003]) and survival studies in transgenic R6/2 and N171-82Q mice. Studies classified as 3 and 4 on our scale would be classified as ‘moderate’ or ‘substantial’ on UK Home Office regulations (Home Office 2000, Richmond J, personal communication; see also Animal Procedures Committee 2003 for a discussion of the UK severity banding). They correspond to the second most severe category, ‘Category D’, used by the Canadian Council on Animal Care (CCAC 1991).
Discussion
When analysing publications of animal models of HD, we found that the majority of experiments were found to be in the two most severe categories in which animals have sensorimotor deficits that interfere with their capacity to eat and drink normally from the cage top. Very few papers reporting such severe experiments included information on refinement measures.
If the reference in scientific publications reflects the actual application of refinement, researchers do not follow the 3Rs principle. We recognize that scientists may apply refinement without referring to it; however, if they do not include such information in publications, it suggests they find it less relevant. Journal publishing policy could play an important role in encouraging scientists to develop and apply refinement techniques, but our results indicate that presently this is not happening. Many journals require authors to state that experimental work with animals was performed in compliance with official regulations, and 3/4 of the papers we reviewed indicated such compliance and/or incorporated general statements referring to the minimization of animal suffering and of the numbers of animals used. However, our study also illustrates that such statements can be less than satisfactory. Blum et al. (2003), for example, state that ‘all efforts to avoid animal suffering were made’, but it is apparent from the paper that this only means ‘as far as the study design already decided upon allowed’. Actually, their approach seems to be considerably more severe than the widely criticized regulatory toxicology test LD 50 which the Organisation for Economic Co-operation and Development (OECD) in 2002 replaced by the fixed dose test applying a more humane endpoint (OECD 2002). Thus, in Blum's first experiment, mice were given daily neurotoxin injections in rising doses until all the animals (n = 36) died; a second experiment was terminated when mortality of around 40% was reached. Similarly, long-term 3-NP administration given as two daily intraperitoneal injections – the method reported in all the reviewed 3-NP studies in mice – cannot be considered the most refined method, given that subcutaneous osmotic minipumps (Anonymous 2007) require only one intervention and allow better dose control (Brouillet et al. 1999).
Our results indicate that the use of death as an endpoint is still relatively common, despite strong advice against this, and international efforts to introduce humane endpoints (Stokes 2000, Demers et al. 2006) which register earlier, less severe clinical signs as endpoint parameters rather than awaiting spontaneous death (e.g. Morton & Hau 2002). In the case of progressive disorders such as the one discussed in this paper, in general, the more the disease is allowed to progress in the experiment, the more severe is the animal harm. Euthanizing animals at an earlier stage will thus ease the amount of distress experienced by each animal. This requires frequent inspections and predefined endpoint criteria related to the disease progress, for example, to euthanize animals when they are found no longer able to right themselves when placed on their back (as used by Dedeoglu et al. 2003). Nevertheless, for most studies, a fixed cut-off point at a certain age would be scientifically satisfactory – and is indeed the approach chosen in the majority of the papers reviewed. Proceeding to late disease stages (where humane endpoints are called for) can only be justified in particular cases, when the objective is to study very specific features of, or the development of treatment for, these stages. Regarding spontaneous death as an endpoint, this is usually preceded by serious sensorimotor disability, and the animals are likely to die from secondary causes such as dehydration or malnutrition (in particular if feed and fluid intake is not facilitated to severely affected animals), which limits the relevance for the study of HD, not at least because the typical human patient at similar disease stages is likely to be in intensive hospital care. For example, in their review paper on the use of R6 mice, Li et al. (2005) questioned that ‘Despite the lack of detailed information on causes of death, it is not uncommon for therapeutic studies to use prolongation of life as one of the main positive indicators of therapeutic efficacy’, further remarking that increased lifespan without improved mobility is of little value to human patients.
The choice of animal model will also have bearings on animal welfare. The transgenic R6/2 mouse model is predominant in the research reports analysed. With the early onset and rapid disease development, the use of this model means that not only experimental animals but also breeders will be affected by considerable sensorimotor deficiencies. From this viewpoint, opting for R6/1 mice can be seen as a refinement; the combination of later onset and slower disease progression makes it possible to follow changes in biological markers without the animals reaching advanced clinical stages (Naver et al. 2003) and animals in reproductive ages retain normal sensorimotor capacity.
An additional observation made during the work with this study is that it is very difficult to find information about clinical symptoms in animals: most papers do not report these and the central databases for ethics evaluations are of limited use, as few countries have searchable databases and existing ones date only a few years back. While shortage of space in scientific journals may be a justification for not describing clinical symptoms unless their evaluation was part of the research, the difficulty of accessing such information may in itself be an ethical problem: an ethics committee that, for the first time, considers an application for work with a model, which has been used previously elsewhere, then have to rely on the applicant's description of the model without being able to benefit from the knowledge gained at other research institutions where the model in question has been studied. The information deficit is even more problematic as regards refinement: adequate description of how refinement such as housing adaptations and humane endpoints were applied would enable personnel at different research institutions to learn from each other in the benefit of animal welfare.
We believe that our study permits two main conclusions to be drawn about the role of scientific journals in enhancing the ethical standards of animal use in research. First, journals should ensure that referees seriously consider whether submitted studies were indeed carried out with the smallest achievable negative impact on the animals involved. In some cases, it may be necessary to involve special referees to deal with this. Secondly, journals should give authors space to describe any measures of refinement and welfare precaution. A positive finding in our study is that some of the reviewed papers did include descriptions of humane endpoints applied (Dedeoglu et al. 2003, Klivenyi et al. 2003, Ferrante et al. 2004, Norflus et al. 2004, Schiefer et al. 2004) or suggested (Naver et al. 2003) and housing adaptations (Hockly et al. 2003, Ribchester et al. 2004); moreover, the five researchers returning questionnaires all referred to refinement measures, although these were not always reported in the publications of their research teams. Overall, however, measures of refinement were rarely reported in research publications. This means that the potential for scientific journals, the main means of communication between researchers, to promote Refinement is poorly utilized. The results of our study give support for the suggestion made by Würbel (2007), that the relevant journals should require that a 3Rs section is included in papers publishing the results of animal-based research.
Footnotes
Acknowledgements
The authors wish to thank colleagues in the COST Action B24 Laboratory Animal Science and Welfare; researchers Nico Dantuna, Christa Maynard, Christoph Kosinski, Elsa Diguet, Dwaine Emerich and Hoa Huu Phuc Nguyen, who provided information on clinical signs in different models; and Paul Robinson, who commented on a previous version of the paper. Preparation of this paper was part of the interdisciplinary project Comparative Genomics of Man and Pig, funded by the Danish Agency for Science, Technology and Innovation.