Most pre-clinical analgesic efficacy assays still involve nociceptive testing in rodents. This is despite concerns as to the relevance of these tests for evaluating the pain-preventative properties of drugs. More appropriate methods would target pain rather than nociception, but these are currently not available, so it remains unknown whether animal pain equates to the negatively affective and subjective/emotional state it causes in humans. Mouse cancer models are common despite the likelihood of substantial pain. We used Conditioned Place Preference (CPP) testing, assessments of thermal hyperalgesia and behaviour to determine the likelihood that MBT-2 bladder cancer impacts negatively on mouse welfare, such as by causing pain. There was no CPP to saline, but morphine preference in tumour bearing mice exceeded that seen in tumour-free controls. This occurred up to 10 days before the study end-point alongside reduced body weight, development of hyperalgesia and behaviour changes. These effects indicated mice experienced a negative welfare state caused by malaise (if not pain) before euthanasia. Due to the complexity of the assessments needed to demonstrate this, it is unlikely that this approach could be used for routine welfare assessment on a study-by-study basis. However, our results show mice in sufficiently similar studies are likely to benefit from more intensive severity assessment and re-evaluation of end-points with a view to implementing appropriate refinements. In this particular case, a refinement would have been to have euthanased mice at least 7 days earlier or possibly by provision of end-stage pain relief. CPP testing was found to be a helpful method to investigate the responses of mice to analgesics, possibly on a subjective level. These findings and those of other recent studies show it could be a valuable method of screening candidate analgesics for efficacy against cancer pain and possibly other pain or disease models.