Preclinical mouse models are critical for understanding the pathophysiological response to infections and developing treatment strategies for sepsis. In keeping with ethical values, researchers follow guidelines to minimize the suffering of the mice. Weight loss is a criteria used as a humane end point, but there is no official recommendation for a maximum weight loss leading to euthanasia. To evaluate whether the thresholds used in daily practice are optimal, we performed a comprehensive retrospective analysis of data generated over 10 years with > 2300 mice used in models of infection with Listeria monocytogenes, Streptococcus pneumoniae, Candida albicans and H1N1 influenza virus. Weight loss segregated mice that survived from those that did not. Statistical analyses revealed that lowering the weight loss thresholds used (none, 30% or 20%) would have increased mortality rates due to the sacrifice of mice that survived infections (p < 0.01-0.001). Power calculations showed high variability and reduction of power as weight loss thresholds approached 20% for S. pneumoniae and L. monocytogenes models. Hence, weight loss thresholds need to be adapted to each model of infection used in a laboratory. Overall, weight loss is a valuable predictor of mortality that contributes to the robustness of composite scores. To our knowledge, this is the most extensive study exploring the relationship between weight loss threshold and sepsis outcome. It underscores the importance of the infection-model-specific evaluation of weight loss for use in clinical scores defining humane endpoints to minimize mouse suffering without compromising statistical power and scientific objectives.