Introduction Whilst alpha feto-protein is secreted by some hepatocellular carcinoma, it is not recommended for diagnosis in current guidelines. It is well recognised as a marker of prognosis in hepatocellular carcinoma and therefore we aimed to determine an appropriate cut-off value that would provide optimal prognostic information for this cancer.

Method Consecutive patients (n = 432) with valid alpha feto-protein measurements diagnosed with hepatocellular carcinoma during 2005 to 2014 in the Liverpool region, UK were included. The area under the receiver operating characteristic curve of alpha feto-protein by status of patients was used to determine the optimal cut-off. This value was then used to assess overall survival of the study population by type of treatment. The predictive performance of the new cut-off was assessed by its performance in the Hepatoma arterial-embolisation prognostic score of patients treated by trans-arterial chemoembolisation.

Results Serum alpha feto-protein value, ≥43 ng/mL predicted prognosis (sensitivity 48% / specificity 78%) better than ≥400 ng/mL (sensitivity 28%/specificity 89%). The median survival was 28 months (95% CI: 21–33) in patients with AFP <43 ng/mL and 7 months (95% CI: 5–8) for those whose alpha feto-protein was ≥43 ng/mL. Also, accurate prediction of survival for hepatocellular carcinoma patients receiving loco-regional therapy (34 months vs. 14 months, p < 0.0001) and curative therapy (43 months vs. 15 months, p = 0.001) was obtained.

Conclusion An alpha feto-protein of ≥43 was defined as the optimal cut-off to aid prognostic predictions. These results require external validation before recommendation for decision making in clinical practice.

Disclosure of interest None Declared.