The diagnostic performance of AFP and PIVKA-II models for non-B non-C hepatocellular carcinoma

Objective This study aims to describe the diagnostic performance of alpha-fetoprotein (AFP), alpha-fetoprotein L3 isoform (AFP-L3), protein induced by vitamin K absence II (PIVKA-II), and combined biomarkers for non-B non-C hepatocellular carcinoma (NBNC-HCC). Results A total of 681 newly-diagnosed primary liver disease subjects (385 non-HCC, 296 HCC) who tested negativity for the hepatitis B surface antigen (HBsAg) and hepatitis C antibody (anti-HCV) enrolled in this study. At the cut-off point of 3.8 ng/mL, AFP helps to discriminate HCC from non-HCC with an area under the curve (AUC) value of 0.817 (95% confidence interval [CI]: 0.785–0.849). These values of AFP-L3 (cut-off 0.9%) and PIVKA-II (cut-off 57.7 mAU/mL) were 0.758 (95%CI: 0.725–0.791) and 0.866 (95%CI: 0.836–0.896), respectively. The Bayesian Model Averaging (BMA) statistic identified the optimal model, including patients’ age, aspartate aminotransferase, AFP, and PIVKA-II combination, which helps to classify HCC with better performance (AUC = 0.896, 95%CI: 0.872–0.920, P < 0.001). The sensitivity and specificity of the optimal model reached 81.1% (95%CI: 76.1–85.4) and 83.2% (95%CI: 78.9–86.9), respectively. Further analyses indicated that AFP and PIVKA-II markers and combined models have good-to-excellent performance detecting curative resected HCC, separating HCC from chronic hepatitis, dysplastic, and hyperplasia nodules.