Are increasing 5-year survival rates evidence of success against cancer?

Context Increased 5-year survival for cancer patients is generally inferred to mean that cancer treatment has improved and that fewer patients die of cancer. Increased 5-year survival, however, may also reflect changes in dia gnosis: finding more people with early-stage cancer, including some who wou ld never have become symptomatic from their cancer. Objective To determine the relationship over time between 5-year cancer sur vival and 2 other measures of cancer burden, mortality and incidence. Design and Setting Using population-based statistics reported by the Nation al Cancer Institute Surveillance, Epidemiology, and End Results Program, we calculated the change in 5-year survival from 1950 to 1995 for the 20 most common solid tumor types. Using the tumor as the unit of analysis, we corr elated changes in 5-year survival with changes in mortality and incidence. Main outcome Measure The association between changes in 5-year survival and changes in mortality and incidence measured using simple correlation coeff icients (Pearson and Spearman). Results From 1950 to 1995, there was an increase in 5-year survival for eac h of the 20 tumor types. The absolute increase in 5-year survival ranged fr om 3% (pancreatic cancer) to 50% (prostate cancer). During the same period, mortality rates declined for 12 types of cancer and increased for the rema ining 8 types. There was little correlation between the change in 5-year su rvival for a specific tumor and the change in tumor-related mortality (Pear son r=.00; Spearman r=-.07). On the other hand, the change in 5-year surviv al was positively correlated with the change in the tumor incidence rate (P earson r=+.49; Spearman r=+.37). Conclusion Although 5-year survival is a valid measure for comparing cancer therapies in a randomized trial, our analysis shows that changes in 5-year survival over time bear little relationship to changes in cancer mortality . Instead, they appear primarily related to changing patterns of diagnosis.