Development and validation of an automated test of animal affect and welfare for laboratory rodents
Lead Research Organisation:
University of Bristol
Department Name: Clinical Veterinary Science
Abstract
A central goal of NC3Rs is the refinement of laboratory procedures to minimise animal pain, suffering, and lasting harm, and to improve laboratory animal welfare. To achieve this, it is essential that we use scientifically validated measures to quantify how experimental, housing, and husbandry procedures affect welfare, and whether refinements do indeed lead to improvements. NC3Rs have therefore called for research to develop new and better welfare measures. Our project will provide such a measure to better assess the welfare of rats and mice (81% of laboratory animals used in the UK in 2010).
The assumption that non-human animals can subjectively experience negative emotional (affective) states and hence suffer underpins concerns for animal welfare. Affective state is thus a key determinant of an animal's welfare. However, because subjective experiences are essentially private, and cannot be measured directly, we need to identify indicators that indirectly reflect an animal's affective state. For example, we can measure physiological changes, such as hormone levels, or behavioural indicators, such as an animal's behaviour in an unprotected open space. However, whilst valuable, such indicators have important limitations which a more recent approach that we have developed seeks to address.
Human psychological studies show that, when feeling happier, people are more likely to judge an ambiguous event as having a positive outcome than when they are feeling down; their proverbial glass is half-full. If the same is true of other animals, then by measuring the judgements they make in ambiguous situations, we may gain useful information about how they are feeling too. To explore this possibility, we have developed a novel, non-verbal test of 'judgement bias' underpinned by theoretical models of decision-making. The test allows us to systematically investigate whether animals judge an ambiguous stimulus positively ('optimistic') or negatively ('pessimistic'). Recent studies of rats, sheep, dogs, monkeys, and starlings support the hypothesis that animals in a more positive affective state do indeed show a more 'optimistic' response to ambiguous stimuli, indicating that this technique may be a very good way of assessing emotional state in animals.
However, existing judgement bias tests are relatively time-consuming, involving hands-on training of animals, and no methods have been developed for mice, the most widely used laboratory species. Therefore, we will develop an automated version of our judgement bias test that can be used for laboratory rodents, including mice. We will then validate this new test by employing a range of carefully-selected pharmacological and environmental treatments, each designed to induce a particular, transient change in affective state. By examining how treatments affect the size of the responses shown, we will be able to see whether the method can quantify the intensity of an emotional state. We will also compare enriched and standard housing conditions, and other determinants of long-term emotional state, to see whether the test can detect the cumulative effects of long-term experience on affect. 'Cumulative suffering' is hard to measure, yet vitally important for an animal's lifetime welfare, and we anticipate that our approach can provide a good indication of this state. An industrial partner has expressed interest in further implementing the test that we develop for widespread use, for example by incorporating it into a home-cage testing system.
The study will also produce computational models of how affective states change in response to external events. These may provide the basis for modelling the effects of experimental treatments on an animal's affective state, hence facilitating better and more humane planning of studies. We hope that this project will lay the foundations for further development of such models.
The assumption that non-human animals can subjectively experience negative emotional (affective) states and hence suffer underpins concerns for animal welfare. Affective state is thus a key determinant of an animal's welfare. However, because subjective experiences are essentially private, and cannot be measured directly, we need to identify indicators that indirectly reflect an animal's affective state. For example, we can measure physiological changes, such as hormone levels, or behavioural indicators, such as an animal's behaviour in an unprotected open space. However, whilst valuable, such indicators have important limitations which a more recent approach that we have developed seeks to address.
Human psychological studies show that, when feeling happier, people are more likely to judge an ambiguous event as having a positive outcome than when they are feeling down; their proverbial glass is half-full. If the same is true of other animals, then by measuring the judgements they make in ambiguous situations, we may gain useful information about how they are feeling too. To explore this possibility, we have developed a novel, non-verbal test of 'judgement bias' underpinned by theoretical models of decision-making. The test allows us to systematically investigate whether animals judge an ambiguous stimulus positively ('optimistic') or negatively ('pessimistic'). Recent studies of rats, sheep, dogs, monkeys, and starlings support the hypothesis that animals in a more positive affective state do indeed show a more 'optimistic' response to ambiguous stimuli, indicating that this technique may be a very good way of assessing emotional state in animals.
However, existing judgement bias tests are relatively time-consuming, involving hands-on training of animals, and no methods have been developed for mice, the most widely used laboratory species. Therefore, we will develop an automated version of our judgement bias test that can be used for laboratory rodents, including mice. We will then validate this new test by employing a range of carefully-selected pharmacological and environmental treatments, each designed to induce a particular, transient change in affective state. By examining how treatments affect the size of the responses shown, we will be able to see whether the method can quantify the intensity of an emotional state. We will also compare enriched and standard housing conditions, and other determinants of long-term emotional state, to see whether the test can detect the cumulative effects of long-term experience on affect. 'Cumulative suffering' is hard to measure, yet vitally important for an animal's lifetime welfare, and we anticipate that our approach can provide a good indication of this state. An industrial partner has expressed interest in further implementing the test that we develop for widespread use, for example by incorporating it into a home-cage testing system.
The study will also produce computational models of how affective states change in response to external events. These may provide the basis for modelling the effects of experimental treatments on an animal's affective state, hence facilitating better and more humane planning of studies. We hope that this project will lay the foundations for further development of such models.
Technical Summary
To assess whether refinements to laboratory procedures do indeed improve animal welfare, it is essential that we use scientifically validated measures of welfare. A key determinant of welfare is an animal's affective (emotional) state. Subjective emotions cannot be measured directly so proxy indicators are used instead. Existing indicators have significant shortcomings and we have therefore developed a new approach, inspired by the links between emotion and cognition observed in humans, and grounded in theoretical models of decision-making. The core hypothesis is that individuals in negative affective states judge ambiguous stimuli negatively compared to happier individuals. We have developed a task to test for such 'cognitive biases' in animals and have found, in several species, that putative affective state is related to judgement bias as predicted. However, our tests are time-consuming and have not yet been extended to mice, the most commonly used laboratory species. We will therefore develop an automated version of our generic judgement bias task for laboratory rodents. We will use environmental and pharmacological manipulations of affective state to validate our new test. We will investigate whether judgement bias size relates to intensity of the induced state (using dose-response studies), and hence whether our approach can quantify emotional intensity. We will also investigate whether the test can detect the cumulative effects of long-term experience on affective state (e.g. by comparing animals kept in enriched and standard housing conditions). An industrial partner has expressed interest in further implementing the test that we develop for widespread use. The study will also produce computational models of how affective states change in response to external events. These may provide the basis for modelling the effects of experimental treatments on an animal's affective state, hence facilitating better and more humane planning of studies.
Planned Impact
This project is in response to an NC3Rs/BBSRC priority highlight call for research into 'animal welfare measures and assessment'. The impact of new and better measures of animal welfare is principally on Refinement, because scientifically validated welfare measures are absolutely essential in order to accurately determine baseline welfare in existing laboratory housing, husbandry and scientific procedures, and to assess whether Refinements really do lead to improvements. New measures of welfare can also encourage Reduction and Replacement of animal use by detecting situations or procedures in which welfare was previously thought to be acceptable, but is now revealed to be very poor and unacceptable. This is most likely if the measures can detect significant welfare problems that may have only subtle overt signs (e.g. depression-like states).
Because there are numerous potential ways in which laboratory procedures may be refined to improve welfare (e.g. altering housing environments to better cater to the animal's needs; minimising pain during experimental studies and marking procedures; changing handling and restraint methods to minimise stress; more accurately detecting when humane end-points should be applied), it is difficult to estimate the numerical impact of better measures of animal welfare on Refinement. However, potentially all animals used in research would benefit from measures which more accurately detect where welfare problems lie, and how effectively they are dealt with by Refinements. In this respect, the impact of a new welfare measure will depend on how well it assesses welfare, whether it can reveal subtle underlying welfare problems, and whether it can be used across species.
We believe that the 'cognitive bias' approach we propose here, inspired by the links between emotion and cognition observed in humans, and based on a novel theoretical framework for assessing animal emotion (affect), which is a key determinant of animal welfare, will score well on all these criteria. Studies of humans and animals indicate that it is a reliable indicator of affective valence (positivity or negativity of emotion - critical to animal welfare), and this will be further validated in our project, as will the sensitivity of the method to varying intensities of affective state. The automated test that we develop will make the method practically viable, and an industry partner (Delta Phenomics BV) has expressed interest in making it more widely available by incorporating it into their behavioural testing systems.
Because the method has a clear theoretical grounding which is generalisable across species, it will be applicable to many different species, and have a potentially wide-ranging impact. It should also be able to reveal affective states that have only subtle overt signs (e.g. depression), and to detect the effects of long-term experience on welfare, or 'cumulative suffering', and hence the impact of housing or other procedures on lifetime welfare. We frequently receive questions from other researchers concerning the cognitive bias approach, and anticipate that an automated method will therefore be of widespread interest to the laboratory animal community.
The mathematical models we develop may provide the foundations for a novel in silico way of predicting how external events impact on affect and welfare, and hence modelling the potential effects of treatments before they are actually used in vivo, thereby facilitating better and more humane planning of studies.
Within our own laboratory, development of an automated cognitive bias test will be of great value in facilitating our use of this measure to evaluate the effects of a range of procedures on laboratory animal welfare. In addition, colleagues studying chronic pain conditions in rodents are already interested in employing the technique to assess cumulative suffering across time, and will collaborate on this project.
Because there are numerous potential ways in which laboratory procedures may be refined to improve welfare (e.g. altering housing environments to better cater to the animal's needs; minimising pain during experimental studies and marking procedures; changing handling and restraint methods to minimise stress; more accurately detecting when humane end-points should be applied), it is difficult to estimate the numerical impact of better measures of animal welfare on Refinement. However, potentially all animals used in research would benefit from measures which more accurately detect where welfare problems lie, and how effectively they are dealt with by Refinements. In this respect, the impact of a new welfare measure will depend on how well it assesses welfare, whether it can reveal subtle underlying welfare problems, and whether it can be used across species.
We believe that the 'cognitive bias' approach we propose here, inspired by the links between emotion and cognition observed in humans, and based on a novel theoretical framework for assessing animal emotion (affect), which is a key determinant of animal welfare, will score well on all these criteria. Studies of humans and animals indicate that it is a reliable indicator of affective valence (positivity or negativity of emotion - critical to animal welfare), and this will be further validated in our project, as will the sensitivity of the method to varying intensities of affective state. The automated test that we develop will make the method practically viable, and an industry partner (Delta Phenomics BV) has expressed interest in making it more widely available by incorporating it into their behavioural testing systems.
Because the method has a clear theoretical grounding which is generalisable across species, it will be applicable to many different species, and have a potentially wide-ranging impact. It should also be able to reveal affective states that have only subtle overt signs (e.g. depression), and to detect the effects of long-term experience on welfare, or 'cumulative suffering', and hence the impact of housing or other procedures on lifetime welfare. We frequently receive questions from other researchers concerning the cognitive bias approach, and anticipate that an automated method will therefore be of widespread interest to the laboratory animal community.
The mathematical models we develop may provide the foundations for a novel in silico way of predicting how external events impact on affect and welfare, and hence modelling the potential effects of treatments before they are actually used in vivo, thereby facilitating better and more humane planning of studies.
Within our own laboratory, development of an automated cognitive bias test will be of great value in facilitating our use of this measure to evaluate the effects of a range of procedures on laboratory animal welfare. In addition, colleagues studying chronic pain conditions in rodents are already interested in employing the technique to assess cumulative suffering across time, and will collaborate on this project.