ObjectiveTo assess the performance of preoperative serum Golgi protein 73 （GP73） in diagnosing liver fibrosis degree in patients with resectable hepatocellular carcinoma （HCC）， to investigate its value in predicting postoperative survival， and to provide a basis for risk stratification and individualized decision-making in this population. MethodsA total of 631 patients with HCC who underwent liver resection at The First Hospital of Jilin University from March 1， 2009 to December 31， 2022 were enrolled. The demographic， pathological， and serological features of the patients were collected， and the serum level of GP73 was measured before surgery. Scheuer score determined by liver biopsy was used as the gold standard for diagnosing liver fibrosis； the Spearman correlation analysis was used to investigate the correlation between GP73 and liver fibrosis； the receiver operating characteristic （ROC） curve and the area under the ROC curve （AUC） were used to assess the value of serum GP73 and compound serological markers in the diagnosis of liver fibrosis， and the Delong test was used for comparison of AUC. With all-cause mortality as the outcome， X-Tile 3.6.1 software was used to determine the optimal cut-off value of GP73； the univariate and multivariate Cox regression analyses were used to identify independent influencing factors for prognosis， and a nomogram model was constructed. Ten-fold cross-validation was used for internal validation； index of concordance （C-index）， integrated discrimination improvement （IDI） index， and net reclassification improvement （NRI） index were calculated； calibration curve and decision curve analysis were used to assess the predictive performance of the model. ResultsThe serum level of GP73 was positively correlated with Scheuer liver fibrosis stage （r_s=0.255， P<0.001）. Serum GP73 had a significantly higher AUC than gamma-glutamyl transferase-to-platelet ratio （0.704 vs 0.634， Z=2.00， P=0.046） and S index （0.704 vs 0.637， Z=1.98， P=0.048） in the diagnosis of significant liver fibrosis， with no significant difference compared with aspartate aminotransferase-to-platelet ratio index （AUC=0.722） and fibrosis-4 index （AUC=0.695） （both P>0.05）. During a median follow-up time of 48.66 months， 239 patients （37.9%） died. GP73 had an optimal cut-off value of 65 ng/mL in predicting postoperative mortality in patients with resectable HCC， and high GP73 expression （≥65 ng/mL） alone achieved a C-index of 0.617 for predicting the prognosis of HCC patients， which was better than the C-index of 0.551 for Barcelona Clinic Liver Cancer （BCLC） stage （P<0.001）. The multivariate Cox regression analysis showed that serum GP73 （hazard ratio ［HR］=2.56， 95% confidence interval ［CI］： 1.86 — 3.52， P<0.001）， sex （HR=1.58， 95%CI： 1.10 — 2.28， P=0.014）， vascular invasion （HR=1.69， 95%CI： 1.30 — 2.21， P<0.001）， total bilirubin （HR=1.49， 95%CI： 1.16 — 1.93， P=0.002）， and maximum tumor diameter （HR=1.07， 95%CI： 1.04 — 1.11， P<0.001） were independent risk factors for all-cause mortality after hepatectomy in patients with HCC. In order to assess whether the addition of GP73 to the other four variables could further improve model performance， the model based on GP73， sex， vascular invasion， total bilirubin， and maximum tumor diameter was defined as the GP73-included model， and the model based on all five variables except GP73 was defined as the GP73-excluded model. The GP73-included model achieved an AUC of 0.704 （95%CI： 0.656 — 0.753） for predicting 5-year survival rate after surgery， and ten-fold cross-validation showed that this model had a mean C-index of 0.716 （95%CI： 0.682 — 0.750）， which was higher than the C-index of the GP73-excluded model （C-index=0.682， 95%CI： 0.645 — 0.719， P=0.008） and BCLC stage （C-index=0.551， 95%CI： 0.523 — 0.580， P<0.001）. Compared with the GP73-excluded model， the GP73-included model yielded an IDI index of 0.039 （95%CI： 0.016 — 0.069， P<0.001） and an NRI index of 0.259 （95%CI： 0.179 — 0.335， P<0.001）， and compared with BCLC stage， the GP73-included model had an IDI index of 0.112 （95%CI： 0.069 — 0.163， P<0.001） and an NRI index of 0.272 （95%CI： 0.169 — 0.380， P<0.001）. Calibration curve and decision curve analyses showed that the GP73-included model had good calibration and net clinical benefit. ConclusionGP73 can be used as a serological biomarker for diagnosing liver fibrosis in patients with resectable HCC， and preoperative GP73 level has a certain value in predicting the survival of these patients after hepatectomy. The integrated prognostic model incorporating GP73 can help to optimize postoperative risk stratification， thereby providing a basis for decision-making in precise treatment of HCC.