Test articles and test systems

To be good "research instruments", animal experiments must be quality assured so they are standardised and reproducible. Quality assurance includes an analysis of the most critical points in an experiment, where any deviations from the protocol could cause major problems. Identifying these points makes it easier to implement measures to prevent the experiment from failing.

Quality assurance also entails characterisation of the "test system" (the animals) and the "test article" (the substance or procedures to which the animals will be exposed). Such characterisation makes it easier to replicate an experiment and obtain the same results in other laboratories. This may be necessary to verify the results or to establish an animal model in another facility. A review in the 1990s of scientific papers describing animal experiments revealed, however, that researchers were generally much better at characterising (and thus standardising) the test article than the test system (Smith et al., 1997). This in itself is not surprising: It is easier to describe the chemicals used in an experiment than the animals. The problem is that by far the greatest potential for variation in an experiment lies within the animals, given their complex biological systems. This variation arises partly due to the animals' intrinsic characteristics (e.g. genotype, health status) and partly due to environmental influences (e.g. temperature, water quality, feed, housing conditions, effects of other animals and humans around them). The situation is still not optimal (Kilkenny et al., 2009). To try to remedy this, several sets of guidelines have been developed that specify how animal experiments should be reported in the literature (Kilkenny et al., 2010; Hooijmans et al., 2010). Many journals have adopted these guidelines, but there are signs that they are not always followed (Baker et al., 2014).

Laboratory animal science is still in its infancy. The learning curve is steep, especially regarding work with those animal species with which we have the least experience. This is particularly true of fish, which make up well over 90% of all laboratory animals in Norway. The varying degrees of "bambi factor" among different species have also influenced the pace of research efforts aimed at improving animal experiments. Recent research showing that fish too have the capacity to feel pain (Sneddon, 2015) has, however, contributed to an increased focus on the welfare of fish species.

Legal, scientific and ethical requirements

The organisation Scand-LAS (http://www.scandlas.org/?page_id=151) has defined laboratory animal science as:
The scientific, legally approved and ethically acceptable study of animals for biomedical purposes.

This definition illustrates that plans for animal experiments must be quality assured on at least three different levels: they must be legal, they must be of high scientific quality and they must be ethical. The last of these is, naturally enough, the most difficult, but scientific standards are not always easy to agree on either. Some focus more on "engineering standards" (e.g. requirements for cage size) while others lean towards "performance standards" (functional requirements, e.g. sufficient space for the animals to thrive). The introduction of new legislation always triggers discussions about the intentions of the law, and how it should be implemented. There are now a great many guidelines, databases, information centres, journals and email discussion groups which provide advice on this subject (3R Guide, 2014).

To weigh up the value of an experiment against the cost to the animal, a harm-benefit analysis must be performed. Such analyses are an integral part of the EU Directive on the protection of animals used for scientific purposes (European Commission, 2010). The Directive came into force in the EU on 1 January 2013, and in Norway the requirements are satisfied via regulations that came into force in 2015 (Ministry of Agriculture and Food, 2015).

A harm-benefit analysis assumes extensive knowledge of the animals' ability to experience discomfort and pain. The development of welfare indicators to measure this scientifically is a rapidly growing field. It is unfortunately easier to identify negative rather than positive indicators. The principle that animals should be given the benefit of the doubt can usefully be applied here. Guidelines have been prepared for classifying the severity of the procedures to which laboratory animals are exposed (Expert Working Group on Severity Classification Criteria, 2009; Hawkins et al., 2011).

The launch of the concept of "the 3 R's" (Replacement, Reduction, Refinement) by Russell & Burch (1959) did much to focus attention on the fact that humane research is also good scientific research. The 3R principle has been incorporated in the legislation on animal testing in many countries: animal experiments should be replaced with alternatives whenever possible, the number of animals should be reduced to an absolute minimum, and experiments that must be performed should be refined to cause the least possible suffering to the animals. The halving of the number of mammals used in experiments in Norway in the 1980s is probably due in part to increased attention to the principle of Reduction (Annual reports from the Norwegian Animal Research Authority http://www.mattilsynet.no/fdu/om/arsrapporter).

A lesser known but very useful set of guidelines when planning animal experiments is the "3 S's" of Carol Newton (1977): Good Science, Good Sense, Good Sensibilities. It must be permissible to use common sense and also to follow one's heart in the absence of scientific data about the harms associated with a procedure.

Striving for Reduction can be a double-edged sword (Hansen et al., 1999). The sum total of suffering is reduced, because pain is experienced by the individual, not the group, but there are two potential dangers: first, it can be tempting to perform too many interventions per animal, and second, the number of animals must not become so low that the experiment loses its ability to deliver statistically significant results. A statistician should without a doubt be involved very early in the planning of animal experiments, to assist in determining the number of animals to be used. Moreover, the development of new statistical methods has shown the potential to more than halve the number of animals needed for certain types of studies without compromising the quality of the results (Dewi et al., 2014).

Another tool that is used increasingly is the determination of humane endpoints. Prior to an animal experiment, a discussion takes place between the researchers, animal technicians and managers of the animal facility to identify criteria for terminating the experiment humanely without requiring the animals to continue until death. There should be an end to "counting bodies on Monday morning".

The objective of an experiment should always be in focus when deciding whether or not it is necessary to use animals. In few areas is this more evident than when animals are considered for use in teaching (Smith & Smith, 2004). Products such as audiovisual aids, three-dimensional models and simulators can be fully valid alternatives to animal experiments for some groups of students. They can even be useful in training persons who will go on to perform animal experiments, by providing a dry run that avoids subjecting living animals to incorrect treatment. Instructors have a statutory duty to familiarise themselves with the options that exist today. Databases are available with information about many thousands of such products (e.g. the NORINA database, http://oslovet.norecopa.no/NORINA).

Naturally there is great political and societal interest in statistics on the number of laboratory animals used in individual countries. In 2013, Norway used more than 5.5 million animals in total for research purposes. But is a large number necessarily a bad sign? Could it just mean that a country has high levels of research activity? Different definitions of animal testing also have an impact. Of the animals used in Norway in 2013, approximately 4.5 million were fish that were used for clinical testing of drugs in commercial farms. The vaccine had already been tested in small-scale laboratory experiments, and the fish were going to be vaccinated anyway. Manufacturers often select a relatively large number of fish farms to ensure that a vaccine's efficacy under different coastal conditions has been adequately examined. Many would argue that these are not laboratory animals in the true sense of the word, and the experiments fall outside the definition of animal experiments in the EU's latest directive on laboratory animals. On the other hand, the new directive has lowered the threshold for procedures that are considered animal experiments to include injections and the equivalent. This means that from 2015 onwards, permission must be sought for some procedures that could previously be performed without approval.

Challenges in housing laboratory animals

A less sensational but equally pertinent example of the many issues around animal testing is the length of time for which animals that neither die nor are killed during an experiment can be kept. Norwegian legislation does not specify maximum housing durations for different animal species. Should animals with long biological lifespans be used for as long as possible? One example is the so-called fistulation of larger domesticated animals such as cows, where a permanent opening is made through the skin to the stomach or intestines, similar to the operation used to attach a colostomy bag in humans. The openings make it possible to extract degradation products from various points in the digestive system, and thus gain a better understanding of where and how nutrients are broken down. Direct pain is mostly limited to a short post-operative phase. Can one thereafter use these animals for years, or should there be a limit to how long they can be used for experiments? Replacing them means performing surgery on new animals, with an additional potential for pain. The "cost" to the animal can to some extent be reduced by measures such as allowing the animal access to pasture.

The optimal housing of laboratory animals should receive careful attention during the planning of an experiment, and must be discussed at an early stage. Both the number of animals per group, and the interactions between the animals and their environment, will have a major impact on factors such as the correct statistical analysis of data from the experiment. Indeed, for the animals themselves, housing conditions can be critical. Housing social animals (e.g. many rodent species and schooling fish) in abnormally small groups will be stressful for them; most animal species will not be happy in a "single room". Recent research suggests that mice housed alone can develop symptoms that in humans would be interpreted as signs of depression (Kalliokoski et al., 2014). However, there will always be a few victims of bullying, who seem relieved to be removed from the group, and it is therefore important to keep an eye on the establishment of the hierarchy.

An increase in the number of laboratory animals in a facility must be accompanied by a proportional increase in human resources to take care of them. The question "How many animals can a technician look after in the course of a week?" should ring alarm bells. Animals must be inspected daily for signs of illness, pain and distress. A professional dilemma is that the animal's requirement for environmental enrichment can make this task more difficult, because they can then hide behind the objects that have been placed in the cage or tank. Animals will also have a need for privacy. This is an especially important consideration for breeding colonies.

Other dimensions (suffering, specification, killing, reuse)

EU Directive 2010/63 (European Commission, 2010) sets an absolute limit in cases where animals are subjected to severe pain over time. How can this be measured? Qualitative scoring systems that provide so-called ordinal data (e.g. on a scale from 1–10) can be devised, but these leave much room for individual interpretation unless every step can be precisely defined. The best option is systems that use continuous data (numbers such as weight and length, which can be placed on a continuous scale). Score sheets are also useful aids (Morton, 1999). Recently, several research groups have developed systems for measuring animals' suffering based on their facial expressions: so-called "Grimace Scales" (e.g. Keating et al., 2012).

Another dilemma occurs when scientists are very specific in their ordering of laboratory animals. This can lead to suppliers increasing their animal breeding in order to fulfil the requests. The number of surplus animals that must be killed will therefore increase and, even though these animals do not show up in a country's statistics, the numbers can be large. These animals will have lived under conditions that are at least as unnatural as those in an animal facility.

Almost all animal experiments end with the animal being killed, whether this is necessary for the experiment or not. Animals in research facilities can end up in lifelong quarantine because of import regulations. Is it unethical to kill these animals (the "badness of killing", Hansen et al., 1999)? The answer will depend in part on one's philosophical viewpoint and evaluation of the moral value of each animal. We do not perform experiments on humans even though it would benefit society, and therefore we should not do so on animals either, many would say. They would also point to the paradox that we justify animal experiments on the grounds that animals show roughly the same responses as humans, and yet at the same time subject animals to treatment that would be completely unacceptable if performed on humans. The method of killing must be discussed carefully as part of the planning of all animal experiments.

Reuse of laboratory animals also creates ethical dilemmas. From what we know today about the memory capacities of animals (including fish), this is potentially a very stressful practice. It can be tempting for several reasons to use animals again, especially in a country like Norway where most laboratory animals are imported. Attacks on airlines have in some cases led to imported animals being subjected to longer transport times through transit countries, because the airlines that fly direct will not carry them. What then is worse: the stress associated with transporting and acclimatising new laboratory animals, or keeping animals longer that are already accustomed to the animal facility? Is being a laboratory animal so far removed from the animal's natural life that an individual's stay in an animal facility should be kept to an absolute minimum, or can we create liveable conditions in which the animals may even thrive?

Animal testing requires a dialogue among all stakeholders in animal research: government bodies, industry, the academic community and animal welfare organisations. In this way, differences of opinion can be discussed in a civilised manner and consensus sought. This will also promote the application of the 3 R's. In Europe, national consensus platforms have been established that work according to this principle. Norway's national platform is called Norecopa (http://www.norecopa.no), and is helping to achieve this goal.

This article has been translated from Norwegian by Louisa Lyon, Akasie språktjenester AS.