Advances and Applications in Statistics
Volume 46, Issue 3, Pages 197 - 223
(September 2015) http://dx.doi.org/10.17654/ADASSep2015_197_223 |
|
ESTIMATING PROSTATE CANCER RELATIVE SURVIVAL AND CANCER-SPECIFIC DEATH
Christopher H. Morrell, Michael W. Kattan, F. Roy MacKintosh, Stephen Van Den Eeden and Thomas B. Neville
|
Abstract: A man’s estimated risk of death from prostate cancer can help guide the selection of the best prostate-specific antigen (PSA) threshold for biopsy. Estimates of all-cause death (ACD) risk are limited to the population studied. Estimates of cancer-specific death (CSD) risk are more useful than ACD risk. There are two methods of estimating CSD risk: (1) using records of the cause of death, and (2) determination of relative survival, a method that removes other causes from ACD risk. Relative survival methods depend heavily on estimates of no-cancer death (NCD) risk rather than on judgment about cause of death; therefore, using relative survival methods to estimate CSD risk can be superior to relying on cause-of-death data based on judgment. Here, we present a novel approach to estimate NCD which allows us to then estimate the CSD risk when cause of death is unavailable. We use prostate cancer data gathered from over 14 million men and 33 million PSA tests at the United States Department of Veterans Affairs (VA). The VA maintains superb death record data for veterans, but lacks good records of cause of death. Therefore, relative survival methods are necessary to estimate CSD risk for VA data, which puts a premium on accurate estimates of NCD risk. |
Keywords and phrases: prostate cancer, relative survival, net survival, Kaplan-Meier, all-cause death, cancer-specific death, no-cancer death. |
|
Number of Downloads: 414 | Number of Views: 1394 |
|