Long-term Cancer Survival Trends by Updated Summary Stage

Cancer Epidemiol Biomarkers Prev. 2023 Nov 1;32(11):1508-1517. doi: 10.1158/1055-9965.EPI-23-0589.

Abstract

Background: Stage is the most important prognostic factor for understanding cancer survival trends. Summary stage (SS) classifies cancer based on the extent of spread: In situ, Localized, Regional, or Distant. Continual updating of staging systems poses challenges to stage comparisons over time. We use a consistent summary stage classification and present survival trends for 25 cancer sites using the joinpoint survival (JPSurv) model.

Methods: We developed a modified summary stage variable, Long-Term Site-Specific Summary Stage, based on as consistent a definition as possible and applied it to a maximum number of diagnosis years, 1975-2019. We estimated trends by stage by applying JPSurv to relative survival data for 25 cancer sites in SEER-8, 1975-2018, followed through December 31, 2019. To help interpret survival trends, we report incidence and mortality trends using the joinpoint model.

Results: Five-year relative survival improved for nearly all sites and stages. Large improvements were observed for localized pancreatic cancer [4.25 percentage points annually, 2007-2012 (95% confidence interval, 3.40-5.10)], distant skin melanoma [2.15 percentage points annually, 2008-2018 (1.73-2.57)], and localized esophagus cancer [1.18 percentage points annually, 1975-2018 (1.11-1.26)].

Conclusions: This is the first analysis of survival trends by summary stage for multiple cancer sites. The largest survival increases were seen for cancers with a traditionally poor prognosis and no organized screening, which likely reflects clinical management advances.

Impact: Our study will be particularly useful for understanding the population-level impact of new treatments and identifying emerging trends in health disparities research.

MeSH terms

  • Humans
  • Incidence
  • Longitudinal Studies
  • Melanoma*
  • Neoplasm Staging
  • SEER Program
  • Skin Neoplasms*