A new approach to the use of α-fetoprotein as surveillance test for hepatocellular carcinoma in patients with cirrhosis.

Background: Surveillance for hepatocellular carcinoma (HCC) is recommended in patients with cirrhosis. As a-fetoprotein (AFP) is considered a poor surveillance test, we tested the performance of its changes over time. Methods: Eighty patients were diagnosed with HCC (cases) during semiannual surveillance with ultrasonography and AFP measurement were recruited and matched for age, gender, etiology and Child-Pugh class with 160 contemporary cancer-free controls undergoing the same surveillance training group (TG). As a validation group (VG) we considered 36 subsequent patients diagnosed with HCC, matched 1:3 with contemporary cancer-free controls. a-Fetoprotein values at the time of HCC diagnosis (T0) and its changes over the 12 (D12) and 6 months (D6) before cancer detection were considered. Results: In both TG and VG, > 80% of HCCs were found at an early stage. In TG, AFP significantly increased over time only in cases. T0 AFP and a positive D6 were independently associated with HCC diagnosis (odds ratio: 1.031 and 2.402, respectively). The area under the curve of T0 AFP was 0.76 and its best cutoff (BC) was 10ngml (sensitivity 66.3%, specificity 80.6%). The combination of AFP > 10ngml or a positive D6 composite a-fetoprotein index (CAI) increased the sensitivity to 80% with a negative predictive value (NPV) of 86.2%. Negative predictive value rose to 99%, considering a cancer prevalence of 3%. In the VG, the AFP-BC was again 10ngml (sensitivity 66.7%, specificity 88.9%), and CAI sensitivity was 80.6% with a NPV value of 90.5%. Conclusions: CAI achieves adequate sensitivity and NPV as a surveillance test for the early detection of HCC in cirrhos