BackgroundCytokeratin is associated with the recurrence and metastasis of some cancers and tends to increase the malignancy of the disease. It is getting more and more attention in cancer research. Abnormal expression of cytokeratin 19 (CK19) has been reported as an important prognostic factor in cancers. CK19 is a marker of bile duct cells, liver progenitor cells (HPCs), and early hepatoblasts, and its expression is associated with poor prognosis in patients diagnosed with hepatocellular carcinoma (HCC). The purpose of this study was to evaluate the predictive value of CK19 for tumor recurrence after radical resection in patients with hepatitis B virus (HBV) positive HCC. MethodsThis study was a retrospective study conducted in two institutions. A total of 674 patients with HBV positive HCC who underwent radical HCC resection from January 2010 to May 2020 were included in this study. Chi-square test or Fisher’s exact test was used to compare the classification variables and continuous variables were compared by t-test or Wilcoxon rank sum test. Cox regression model was used for univariate and multi-variable survival analyses. Based on the results of the multi-variable analyses of Cox regression, the nomogram of 2-year recurrence-free survival (RFS) was plotted. The model was validated internally in the Hangzhou cohort (training set) and then externally in the Lanzhou cohort (test set) and the effectiveness of the model was tested. ResultsFor all 674 patients, 223 cases (33.1%) were positive and 451 cases (66.9%) were negative for CK19. The 2-year RFS rate was higher in patients with CK19 negative than in patients with CK19 positive. In the training set, correlation analysis showed that CK19 expression was correlated with preoperative potassium (P value(P) = 0.030), satellite nodules (P < 0.001) and microvascular invasion (P = 0.020). In the test set, CK19 expression was correlated with postoperative platelet (P = 0.038), satellite nodules (P = 0.003), microvascular invasion (P = 0.011), and maximum tumor size (P = 0.039). Univariate Cox regression correlation analyses showed that CK19 expression was correlated with preoperative potassium (P value(P) = 0.030), satellite nodules (P < 0.001), and microvascular invasion (P = 0.020). Training and test sets showed that postoperative platelet (> 300/L), CK19, satellite nodules in the training set, microvascular invasion, maximum tumor size, and tumor boundary were adverse factors for predicting RFS. Multi-variable analyses showed that in the training set, postoperative platelet > 300/L (hazard ratios (HR) = 2.753, 95% confidence interval (95%CI):1.234–6.142, P = 0.013), CK19 (HR = 1.410, 95%CI:1.006–1.976, P = 0.046), satellite nodule (HR = 1.476, 95%CI:1.026–2.120, P = 0.036), microvascular invasion (HR = 2.927, 95%CI:2.006–4.146, P < 0.001), incomplete tumor capsule (HR = 1.539, 95%CI:1.012–2.341, P = 0.044) were independent prognostic indicator of poor RFS. In the test set, postoperative platelet > 300/L (HR = 2.816, 95%CI:1.043–7.603, P = 0.041), CK19 (HR = 1.586, 95%CI:1.016–2.475, P = 0.042), satellite nodule (HR = 1.706, 95%CI:1.067–2.728, P = 0.026), microvascular invasion (HR = 1.611, 95%CI:1.034–2.510, P = 0.035), and tumor without capsule (HR = 1.870, 95%CI:1.120–3.120, P = 0.017) were independent prognostic indicators of poor RFS. The C-index for the nomogram was 0.698 (95%CI: 0.654–0.742) and the C-index for the test set was 0.670 (95%CI: 0.616–0.724). Both internal and external verification showed good results in identification and calibration. ConclusionCK19 plays a key role in tumor malignancy through overexpression and the expression of CK19 is an independent adverse factor affecting recurrence; therefore, CK19 can be used as a potential biomarker to predict adverse prognosis after surgery and adjuvant therapy in HCC patients.