Background: It is not established whether virological status affects the efficiency of alpha-fetoprotein (AFP) as a hepatocellular carcinoma (HCC) marker among patients with chronic liver disease (CLD).
Methods: We enrolled in a case-control study 170 HCC and 170 CLD patients, matched for age, sex, CLD and HBsAg/anti-HCV status. The AFP sensitivity, specificity, positive (PPV) and negative (NPV) predictive values were calculated. PPV and NPV were evaluated for three additional HCC prevalences (5, 10, and 20%).
Results: The best discriminating AFP value was 16 ng/ml. A value of 20 ng/ml (above which investigations for HCC are recommended) had equivalent sensitivity (60.0 vs. 62.4%) and specificity (90.6 vs. 89.4%). PPV of 20 ng/ml was 84.6% but decreased to 25.1% at 5% tumor prevalence. NPV was 69.4% and rose to 97.7% at 5% prevalence. In the different groups of infected patients PPV ranged from 80.0 to 90.9%, falling to 17.4-34.5% at 5% prevalence. In noninfected patients PPV was 100% at any HCC prevalence. NPV ranged from 59.0 to 73.0%, reaching 96.5-98.1% at 5% prevalence.
Conclusions: In CLD patients, AFP monitoring misses many HCCs and inappropriately arouses suspicion of malignancy in many patients. Its usefulness is barely affected by the infection responsible for CLD. An AFP elevation could be more indicative of HCC in non-infected patients.