A Semiparametric Model of Personal Cure

• Activity Number: 255
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Margaret Stedman and Joanne Chang and Kathleen Cronin and Angela Mariotto*+
• Companies: National Cancer Institute and National Cancer Institute and National Cancer Institute and National Cancer Institute
• Keywords: personal cure ; cancer ; survival ; SEER data ; competing risk ; cure models

Abstract

Personal cure is the percent of cancer patients that will die of other causes unrelated to their cancer diagnosis (in the presence of other competing causes of death). It differs from biological cure, the percent of cancer patients that will never die of their cancer (in the absence of other competing causes of death). In this paper we introduce a semi-parametric method for estimating personal cure using independent data sources. Net cancer survival is estimated from a mixture cure model applied to SEER population-based cancer survival data. Net other-cause survival is estimated from US life tables matched to the cancer patient's age, sex, race, and year. Assuming independent competing causes, personal cure is estimated as the crude probabilities of other-cause death. In an example of colorectal cancer survival data, we demonstrate that young men aged 45-54 have lower personal cure rates (69%) than older men aged 65-74 (72%), while the estimate of biological cure is 63% across age groups. Crude measures such as personal cure take into account the increased burden of co-morbidity in the older population to provide a more representative measure of survival.