Sa1866 Alpha-Fetoprotein (AFP) in Hepatocellular Carcinoma (HCC): Determination of Optimal Cut Off for Prognosis Prediction

Abstract

Background Whilst AFP is secreted by some HCCs, it is not recommended as a test in current guidelines owing to its low sensitivity. AFP is felt to provide prognostic information. The cut value for this purpose has been arbitrarily based on a previous diagnostic cut off of 400ng/mL. We aim to determine the optimal cut value for AFP that could achieve prognostic utility. Method Consecutive HCC patients (n=469) with a valid AFP measurement diagnosed during 2005 to 2014 in the Liverpool region, UK were included. Area under the receiver operating characteristic curve (ROC) was constructed to define optimum cut values to achieve high sensitivities by bootstrapping and likelihood ratio (LR) methods. 1000 iterations were involved in determining the value at fixed sensitivities and specificities with the best predictor used to assess overall survival (OS) for those above and below the cut value. This process was repeated for LR. The cut value with the optimum accuracy was used to assess OS for recruited subjects; relative to treatment type [supportive care (BSC), systemic therapy, loco-regional therapy and curative therapy] using Kaplan Meier survival statistic. The performance of the new cut was assessed by comparison of ROC of Hepatoma arterial-embolisation prognostic (HAP) score (using AFP>400ng/mL) versus the modified score (using new cut off) in patients treated with transarterial chemoembolization (TACE). Results The serum AFP cut value by LR (>43ng/mL) predicted prognosis better (sensitivity 48% / specificity 78%) than the cut of >400ng/mL (sensitivity 28% / specificity 89%). The median survival was 28 months (95% confidence interval: 21-33) in patients with AFP 43ng/mL. This lower AFP cut accurately predicted OS for HCC patients receiving loco-regional therapy (34 months vs. 14 months, p 43 was higher than HAP score-AFP>400 (0.7 vs 0.6, p=0.17).Conclusion AFP value of 43 ng/mL was defined as a cut to determine prognosis in this group of patients overall and specifically those receiving potentially curative or loco-regional therapy. These results require validation in a separate cohort of patients to determine whether the cut of 43 can be used to influence decision making in clinical practice.