Cancer is a leading cause of death in domestic dogs. Deaths due to cancer vary widely among breeds, providing an opportunity for testing the multi-stage model of carcinogenesis. This model underpins evolutionary and basic studies of cancer suppression and predicts a linear increase in cancer with breed size, an expectation complicated by bigger breeds having a shorter lifespan (decreasing risk). Using three independent datasets, the weight and lifespan of breeds provided a good fit of lifetime cancer mortality to the multi-stage model, the fit suggesting many canine cancers are initiated by four driver mutations. Of 85 breeds in more than one dataset, only flat-coated retriever showed significantly elevated cancer mortality, with Scottish terrier, Bernese mountain dog and bullmastiff also showing notable risk (greater than 50% over expected). Analysis of breed clades suggested terriers experience elevated cancer mortality. There was no evidence that the lower mass-specific metabolic rate of larger breeds reduced cancer risk. Residuals indicated increased breed inbreeding shortened expected lifespan, but had no overall effect on cancer mortality. The results provide a baseline for identifying increased breed risk for specific cancers and demonstrate that, unless selection promotes increased cancer suppression, the evolution of larger longer-lived animals leads to a predictable increased cancer risk.