Background: Combined hepatocellular-cholangiocarcinoma (cHCC-CCA) is poorly understood, while the predictive value of the staging in which it is included is controversial. Methods: Patients with cHCC-CCA underwent radical hepatectomy in two medical centers in China were enrolled and staged based on optimal cut-off values of tumor burden score (TBS), determined using the X-Tile. The association between TBS and prognosis was evaluated by Cox proportional hazard models and Kaplan-Meier curves with Log-rank test. TBS model and primary liver cancer (PLC) stages were compared by discrimination, consistency, and clinical utility, which were further validated by a 5-folds cross-validation. Results: A total of 192 patients were stratified into low, medium, and high TBS, comprising 92, 51 and 49 patients, respectively. Prognoses worsened with elevated TBS in both the training and validation cohorts. TBS was not only an independent prognostic indicator in univariate and multivariate cox regression, but also a stable risk factor in subgroup analysis according to baseline variables. TBS exhibited best discrimination within these predictive models, as evidenced by the highest c-index and area under curve values of time-dependent receiver operating curves within 5 years post-surgery. TBS calibration plots revealed favorable consistency between prediction and observation. Decision curve analysis suggested higher net benefits for TBS. A 5-folds crossvalidation revealed consistent results. Conclusions: TBS could be applied to stratify cHCC-CCA patients after surgery into groups with statistically different prognoses. Moreover, TBS exhibited optimal prognostic value over all available PLC stages and may inform clinical decisions.