Estimating Personal Cure in Colorectal Cancer Patients Using the SEER Data

• Activity Number: 479 - Survival Analysis II
• Type: Contributed
• Date/Time: Wednesday, August 1, 2018 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Margaret Stedman*
• Companies: Stanford Univ
• Keywords: Survival analysis; Cure models; Cancer Statistics; SEER data; lifetables; prediction

Abstract

Personal cure is a statistic used to predict the long-term probability of dying of non-cancer related causes given the lifespan of the population. Current methods rely on adequate follow-up time and give outdated results. We propose a semi-parametric approach to modeling personal cure which combines the predicted cancer survival from the parametric mixture cure model with stratified rates of all-cause mortality from the US life tables (non-parametric). A competing risk model is used to join results from the two data sources. 2007 predictions from SEER colorectal cancer data were 83-89% for localized stage, 66-72% for regional stage, and 12-14% for distant stage. Accuracy of the mixture cure model was improved with the addition of the joinpoint method, however distant stage cure rates are likely underestimated due to less follow-up data. The probability of personal cure summarizes in one measure the risk of dying of cancer and dying of other causes making it an important statistical tool for clinical decision making.