Prognostic impact of recorded bone, brain, and lung metastatic patterns in hepatocellular carcinoma: a population-based cohort study using the SEER database

Abstract This retrospective cohort study identified patients diagnosed with HCC between 2010 and 2022 who had at least one recorded bone, brain, or lung metastasis at initial diagnosis in the Surveillance, Epidemiology, and End Results (SEER) database. Metastatic patterns were categorized by the number of recorded sites (one vs. multiple) and by organ involvement (bone, brain, or lung). Kaplan-Meier curves, Cox proportional hazards models, and cause-specific and Fine-Gray competing risk models were used to analyze associations of metastatic patterns with overall survival (OS) and with the cumulative incidence of HCC-attributable and other-cause death. A total of 4,282 patients were included, comprising 3,675 with one recorded site and 607 with multiple recorded sites, and 3,932 deaths occurred, including 3,619 HCC-attributable, 252 other-cause, and 61 unknown-cause deaths. Multiple recorded sites were independently associated with worse OS than one recorded site (HR = 1.221, 95% CI: 1.115–1.337, p < 0.001) and with a higher cumulative incidence of HCC-attributable death (SHR = 1.269, 95% CI: 1.160–1.388, p < 0.001), whereas the subdistribution hazard of other-cause death was lower (SHR = 0.508, 95% CI: 0.321–0.804, p = 0.004). Among patients with one recorded site, lung metastasis was associated with higher mortality than bone metastasis (HR = 1.194, 95% CI: 1.101–1.295, Holm-adjusted p < 0.001), whereas the brain versus bone (HR = 1.265, 95% CI: 0.955–1.675) and brain versus lung (HR = 1.059, 95% CI: 0.798–1.406) comparisons were not significant. Later year of diagnosis, primary-site surgery, radiotherapy, and chemotherapy were associated with lower hazards of death, and these associations should not be read as causal treatment effects. Both the number and the location of recorded metastatic sites are associated with prognosis in HCC with bone, brain, or lung metastasis at diagnosis. Multiple recorded sites were associated with both higher all-cause mortality and a higher cumulative incidence of HCC-attributable death, so the excess risk is not explained by non-cancer death. Reporting cumulative incidence for both causes avoids the bias introduced by censoring competing events and supports realistic prognostic counselling.