Previous investigators argued that increasing 5-year survival for cancer patients should not be taken as evidence of improved prevention, screening, or therapy, because they found little correlation between the change in 5-year survival for a specific tumor and the change in tumor-related mortality. However, they did not control for the change in incidence, which influences mortality and is correlated with 5-year survival. The purpose of this study was to reexamine the question of whether increasing 5-year survival rates constitute evidence of success against cancer. We estimate the relationship across cancer sites between long-run changes in population-based mortality rates and both (1) changes in 5-year relative survival rates, and (2) changes in incidence rates, using data from both the U.S. and Australia. We analyze two outcome measures, and the relationship between them: the unconditional mortality rate (number of deaths per 100,000 population), and the 5-year relative survival rate. When incidence growth is controlled for, there is a highly significant correlation, in both the U.S. and Australia, between the change in 5-year survival for a specific tumor and the change in tumor-related mortality. The increase in the relative survival rate is estimated to have reduced the unconditional mortality rate by about 15% in the U.S. between 1976 and 2002, and by about 15% in Australia between 1984 and 2001. While the change in the 5-year survival rate is not a perfect measure of progress against cancer, in part because it is potentially subject to lead-time bias, it does contain useful information; its critics may have been unduly harsh. Part of the long-run increase in 5-year cancer survival rates is due to improved prevention, screening, or therapy.