Generalizability, Robustness and Replicability When Evaluating Wellbeing of Laboratory Mice with Various Methods

SIMPLE SUMMARY: It is in the interest of the general public as well as the scientific community to optimize the wellbeing of animals during scientific research. To reach this goal, methods need to be defined which can reliably evaluate the wellbeing of animals. In this study, we assessed whether various methods, such as measuring body weight, burrowing activity, nesting behavior, a distress score and fecal corticosterone metabolites can differentiate between healthy mice and mice after surgical intervention or during the progression of a gastrointestinal disease. The ability of each method to differentiate between these two states of wellbeing was different between distinct surgical interventions and gastrointestinal diseases. These data suggest that scientists cannot rely on a single method, but have to combine many methods when assessing the wellbeing of animals. ABSTRACT: An essential basis for objectively improving the status of animals during in vivo research is the ability to measure the wellbeing of animals in a reliable and scientific manner. Several non-invasive methods such as assessing body weight, burrowing activity, nesting behavior, a distress score and fecal corticosterone metabolites were evaluated in healthy mice and after three surgical interventions or during the progression of four gastrointestinal diseases. The performance of each method in differentiating between healthy and diseased animals was assessed using receiver operating characteristic curves. The ability to differentiate between these two states differed between distinct surgical interventions and distinct gastrointestinal diseases. Thus, the generalizability of these methods for assessing animal wellbeing was low. However, the robustness of these methods when assessing wellbeing in one gastrointestinal disease was high since the same methods were often capable of differentiating between healthy and diseased animals independent of applied drugs. Moreover, the replicability when assessing two distinct cohorts with an identical surgical intervention was also high. These data suggest that scientists can reach valid conclusions about animal wellbeing when using these methods within one specific animal model. This might be important when optimizing methodological aspects for improving animal wellbeing. The lack of generalizability, however, suggests that comparing animal models by using single methods might lead to incorrect conclusions. Thus, these data support the concept of using a combination of several methods when assessing animal welfare.