Time-varying prognostic impact of AJCC staging on overall survival and cancer-specific survival in hepatocellular carcinoma: a conditional survival analysis using population-based data

The prognostic association of AJCC stage may change with elapsed survival. We evaluated this association in patients with hepatocellular carcinoma (HCC). We analyzed 21,492 HCC patients diagnosed in 2004–2022 from the Surveillance, Epidemiology, and End Results (SEER) database, classified according to the AJCC seventh edition. Time-varying Cox models with restricted cubic splines (RCS) and landmark analyses evaluated overall survival (OS) and net cancer-specific survival (CSS) at years 0–5. Primary models adjusted for age, sex, race, and area-level income. Prediction models underwent internal validation using a random 7:3 split, assessing discrimination, calibration, and exploratory decision curves. Competing risk and sensitivity analyses were also performed. Stage-associated HRs attenuated over time. Five-year conditional OS for stage IV increased from 3.6% (95% CI 3.0–4.4%) at baseline to 47.5% (33.9–59.9%) among five-year survivors. Adjusted HRs for stages III/IV versus stage I decreased by year 5, although stage IV remained associated with higher mortality. At the year-5 landmark, AUCs ranged from 0.625 to 0.694 for OS and from 0.623 to 0.703 for CSS. Mean predicted mortality risks were close to, but slightly lower than, observed risks. A web-based conditional survival calculator was developed. While a higher initial AJCC stage is associated with greater early mortality, its prognostic influence wanes over time. Conditional survival estimates may support real-time risk reassessment by incorporating time already survived. External validation is required before these models can guide clinical decisions.