Ambiguities in Calculating Cancer Patient Survival

When estimating survival from cancer registry data, there are several decisions to be made concerning record selection and method of calculation. For instance, should one use relative or disease-specific survival and how should one handle 'autopsy only' cases and cases with unknown cause of death. This study evaluates the effects of such decisions for survival from colorectal and prostate cancer in the SEER program. The research population consists of cases of colorectal cancer and prostate cancer as reported to a SEER cancer registry from 1973 through 1994. Several alternative methods of estimating survival were evaluated. One method of disease-specific survival is chosen as index scenario, and variants in the selection of cancer cases and the method of estimating survival are compared to this index scenario. In general, the differences between the different survival estimates are small for colorectal cancer and somewhat larger for prostate cancer. The most substantial differences concern disease-specific versus relative survival for prostate cancer and inclusion versus exclusion of individuals with more than one cancer for both colorectal and prostate cancer. There is no single best method for calculating cancer survival. Nevertheless, suggestions for calculating cancer survival when using SEER program data are derived from this study. MESH HEADINGS Bias; Cancer; Colorectal Neoplasms; Epidemiologic Methods; Prostatic Neoplasms; SEER Program; Survival Rate;