Reduction of animal suffering during in vivo experiments is usually ensured by continuously monitoring the health status using a score sheet and by applying humane endpoints. However, most studies do not evaluate the plausibility of score sheets and do not attempt to reduce the suffering of animals by determining earlier and, therefore, more humane endpoints. The present study uses data from BALB/cANCrl mice after bile duct ligation to retrospectively analyze which score sheet criteria are informative to determine humane endpoints. The performance of each single as well as com­binations of multiple animal welfare parameters was analyzed by a Cox proportional-hazards model followed by Harrell’s concordance index. The addition of behavioral parameters, such as burrowing activity, helped to define a more humane early endpoint for euthanizing these animals. Using this approach, we determined that a body weight loss of 10-20% combined with a reduction of burrowing activity by more than 79.4% was able to predict that these animals would die within two days. Thus, this approach successfully determined an earlier humane endpoint and will reduce the suffering of animals in future experiments. Application of such an approach or similar methods can contribute to the refinement of various animal experiments.