A prospective cohort validation study of a multivariable algorithm ASAP for risk stratification of surveillance for hepatocellular carcinoma

INTRODUCTION AND OBJECTIVES: ASAP is a multivariable risk model for hepatocellular carcinoma (HCC). This prospective pilot study evaluated its performance, based on a previously established cut-off value, in predicting HCC risk in patients with chronic liver disease and compare it to alpha-fetoprotein (AFP) and protein induced by vitamin K absence or antagonist-II (PIVKA-II) alone. PATIENTS AND METHODS: This study followed prospective-specimen collection and retrospective-blinded-evaluation (PRoBE) guidance for a phase III biomarker study. Patients undergoing 6-monthly HCC surveillance with ultrasound and AFP were enrolled between December 2017 and October 2018. Serum samples for AFP and PIVKA-II were obtained at surveillance and ASAP scores were calculated. Kaplan-Meier and Cox regression models were used to evaluate the predictive performance of ASAP algorithm and biomarkers for HCC development. RESULTS: Of 612 patients enrolled, 15 developed HCC (all early stage) during a median follow-up of 52.2 months. Using a two-tier risk approach, ASAP model, AFP and PIVKA-II were able to stratify risk of HCC development over time. ASAP low-risk group had the lowest 12-month HCC cumulative risk at 0.18%, versus PIVKA-II at 0.19%, and AFP at 0.51%. Cox modelling demonstrated that increased ASAP score was associated with higher risk of HCC (HR 1.80 for 10% increase, p<0.001), with ASAP providing the highest C-index of 0.7706 versus AFP (0.7285) and PIVKA-II (0.6840). CONCLUSIONS: In this pilot study, the ASAP model demonstrated high accuracy in stratifying the risk of HCC development in patients. It can potentially identify high-risk patients while sparing low-risk individuals from unnecessary imaging surveillance.