Stage harmonized temporal survival comparisons in de novo metastatic breast cancer using SEER

Abstract Temporal survival comparisons in cancer registries depend on how stage is harmonized across changing coding systems. We analyzed SEER 17 breast cancers diagnosed in 2000–2023 using a year-specific TNM-priority M-stage algorithm. The primary cohort comprised 35,262 de novo M1 cases diagnosed in 2010–2023 with complete covariate and outcome data. Contemporary (2014–2023) versus Transition (2010–2013) diagnosis was associated with lower breast cancer-specific mortality before adjustment (hazard ratio [HR], 0.870; 95% confidence interval [CI], 0.845–0.897) and after demographic and organ-site adjustment (HR, 0.844; 95% CI 0.819–0.870). Transition-target inverse probability weighting produced a similar estimate (HR, 0.853; 95% CI 0.827–0.879). A strict TNM analysis was unchanged, whereas the prior any-positive-stage rule yielded a near-null crude estimate. Because proportional hazards were not constant, the overall HR is an average association; adjusted HRs were 0.879 during 0–24 months and 0.801 after 24 months. The 36-month overall-survival restricted mean difference was 0.66 months (95% CI 0.34–0.99). These observational findings show that stage harmonization materially influences temporal comparisons and do not identify effects of imaging or treatment.