Abstract

Background: The thresholds of alpha-fetoprotein (AFP) and des-gamma-carboxy prothrombin (PIVKA-II) when detecting hepatocellular carcinoma (HCC) in chronic hepatitis B (CHB) patients with antiviral nucleoside analog (NA) remain controversial. A relevant integrated nomogram needs to be developed.
Methods: We enrolled a consecutive series of 5666 cases diagnosed with CHB either with or without antiviral agents and randomly allocated them to the training set (n=3966, 70.00%) and the validation set (n=1700, 30.00%).
Results: In the training set, the levels of AFP and PIVKA-II of NA-treated patients were significantly lower than those of untreated patients. The most appropriate cut-off values of AFP and PIVKA-II were 151.40 ng/mL (a sensitivity of 39.77% and a specificity of 92.17%) and 35.50 mAU/mL (a sensitivity of 84.85% and a specificity of 69.43%) for NA-treated patients. As for BCLC-0/A HCC, the most appropriate cut-off values of AFP and PIVKA-II were 151.40 ng/mL and 32.50 mAU/mL for NA-treated patients, respectively. A logistic regression model composed of AFP, PIVKA-II and other clinical parameters to predict the risk of HBV-related HCC for NA-treated patients was established and verified to have an AUROC of 0.868 (95% CI, 0.827– 0.909) for all-stage HCC and an AUROC of 0.856 (95% CI, 0.809– 0.903) for BCLC-0/A HCC.
Conclusion: The new detection thresholds of AFP and PIVKA-II might lead to the ability to perform early detection for hepatoma in NA-treated patients and the innovative risk prediction model is a valuable tool for identifying high-risk CHB patients.

Keywords: hepatitis B virus, hepatocellular carcinoma, alpha-fetoprotein, des-gamma-carboxy prothrombin, antiviral agents, nomogram