Abstract
PURPOSE: Early detection of hepatocellular carcinoma (HCC) through surveillance could reduce this cancer-associated mortality. We aimed to develop and validate algorithms using panel serum biomarkers to identify HCC in a real-world multi-center study in China. PATIENTS AND METHODS: A total of 10,359 eligible subjects, including HCCs and benign liver diseases (BLDs), were recruited from six Chinese medical centers. The three nomograms were built using logistic regression and their sensitivities and specificities were carefully assessed in training and validation cohorts. HCC patients after surgical resection were followed to determine the prognostic values of these algorithms. Prospective surveillance performance was assessed in a cohort of chronic hepatitis B patients during 144 weeks follow-up. RESULTS: Independent risk factors such as alpha-fetoprotein (AFP), lens cuinaris agglutinin-reactive fraction of AFP (AFP-L3), des-gamma-carboxy prothrombin (DCP), albumin (ALB), and total bilirubin (TBIL) obtained from train cohort were used to construct three nomograms (LAD, C-GALAD, and TAGALAD) using logistic regression. In the training and two validation cohorts, their AUCs were all over 0.900, and the higher AUCs appeared in TAGALAD and C-GALAD. Furthermore, the three nomograms could effectively stratify HCC into two groups with different survival and recurrence outcomes in follow-up validation. Notably, TAGALAD could predict HCC up to 48 weeks (AUC: 0.984) and 24 weeks (AUC: 0.900) before clinical diagnosis. CONCLUSION: The proposed nomograms generated from real-world Chinese populations are effective and easy-to use for HCC surveillance, diagnosis, as well as prognostic evaluation in various clinical scenarios based on data feasibility.