Discrepancies in cancer mortality estimates.

Progress against cancer-associated mortality could be estimated by mortality rates. However, population-based measures may not be comparable to trends across time based on individual cases followed over time. For two leading cancer types (colorectal and lung and bronchus), the following were calculated from the Surveillance, Epidemiology, and End Results (SEER) Program survival matrix output: probability of death (PD) as the relative cumulative deaths in a cohort of cancer patients at 12-month intervals after the year of diagnosis, 5-year survival probability, and median survival time. Annual age-adjusted U.S. mortality rates were obtained from the National Center for Health Statistics (NCHS). Colorectal cancer PD 5-year (patients having survived up to 5 years after diagnosis) decreased 4.3% from 1985-1997, in contrast to a 20% decrease in the mortality rate during the same period. The mortality rate for lung and bronchus cancer sharply increased between 1973 and 1991 and was followed by a clear downward trend of 2.5% from 1991-1997, while its PD 5-year decreased by only 0.79%. PD for specific cohorts at yearly intervals did not improve across time at comparable intervals; however, the mortality rate was clearly reduced for both anatomic sites. Survival and PD were, as expected, inversely related. Changes in median survival and 5-year survival probability were similar over time. Although the cancer mortality rate is a clear statistical end-point, inference of success by its falling trend alone could be unrelated to the SEER probability of death.