Chinese Hepatolgy ›› 2026, Vol. 31 ›› Issue (2): 228-233.

• Liver Tumor • Previous Articles     Next Articles

Construction and comparison of hepatocellular carcinoma prediction models based on serum AFP, AFP-L3 and PIVKA-Ⅱ

LI Bo, WANG Hai-yu, JIANG Wen, YU Jin-hong, ZHANG Yong-chen   

  1. 1. Department of Laboratory Medicine, the Second Hospital of Nanjing, Nanjing Hospital Affiliated to Nanjing University of Chinese Medicine, Nanjing 210003, China;
    2. Department of Oncology, the Second Hospital of Nanjing, Nanjing Hospital Affiliated to Nanjing University of Chinese Medicine, Nanjing 210003, China
  • Received:2025-02-06 Online:2026-02-28 Published:2026-04-17
  • Contact: ZHANG Yong-chen,Email:zhangyongchen@126.com

PDF (PC)

1

Abstract: Objective To construct a prediction model for hepatocellular carcinoma (HCC) based on serum alpha fetoprotein (AFP), alpha fetoprotein isoform (AFP-L3), and protein induced by vitamin K absence or antagonist-Ⅱ (PIVKA-Ⅱ), and to compare its value with previously published predictive models, exploring its clinical application significance. Methods A total of 106 patients with HCC and 60 patients with liver cirrhosis were collected from the Second Hospital of Nanjing between January and May 2023 as the training set. An additional 40 cases of HCC and 46 cases of liver cirrhosis were collected from May to July 2024 as the validation set. A risk prediction model for HCC was established using stepwise backward logistic regression, and its efficacy was validated in the validation set. The diagnostic value of the model was evaluated by comparing the area under the receiver operating characteristic (ROC) curve (AUC), sensitivity, specificity, and accuracy with two previously reported HCC predictive models. Results The levels of AFP [596.85 (60.71, 16 338.00)(ng/mL)], AFP-L3 [58.07 (7.05,2730.50)(ng/mL)]and PIVKA-Ⅱ [1 934.50 (59.42, 19 036.25)(ng/mL)] in HCC group were significantly higher than those in cirrhosis group, and the differences were statistically significant (P<0.01). Through multivariate logistic regression analysis, AFP [OR(95%CI): 1.030(1.008~1.052), P=0.007], AFP-L3 [OR(95%CI): 0.971(0.950~0.992), P=0.007] and PIVKA-Ⅱ [OR(95%CI): 1.002(1.000~1.004), P=0.031] were identified as predictive variables for HCC diagnosis. A model was constructed based on these variables. The AUC for the ROC curve in the training set was 0.979 (95%CI: 0.943~0.995), with sensitivity and specificity of 94.30% and 91.70%, respectively. In the validation set, the AUC was 0.939 (95%CI: 0.866~0.979), with sensitivity and specificity of 100% and 76.10%, respectively. Our constructed model demonstrated good diagnostic performance compared to the ASAP and GALAD models. Conclusion The developed predictive model effectively assesses the risk of HCC in high-risk patients, providing important clinical evidence for early diagnosis of hepatocellular carcinoma.

Key words: Alpha fetoprotein, Alpha fetoprotein isoform, PIVKA-Ⅱ, Hepatocellular carcinoma, Predictive model