Alpha-fetoprotein testing for hepatocellular carcinoma may not be helpful in nonalcoholic steatohepatitis

Background & Objectives: Diagnosing hepatocellular carcinoma (HCC) often utilizes serum tumor markers. Although the most commonly used tumor marker in clinical practice, alpha-fetoprotein (AFP) is not included in recent guidelines for diagnosing HCC. The overall performance characteristics of AFP as a tumor marker is viewed as insufficiently sensitive or specific. The diagnostic value of AFP specifically in nonalcoholic steatohepatitis (NASH) related HCC is unknown. We aimed to determine the utility of AFP testing in NASH-related HCC. Methods: Retrospective review of 737 HCC patients referred from 19932011 to a single facility treating the majority of chronic liver disease in Hawaii. HCC was diagnosed histologically by percutaneous biopsy, liver biopsy at the time of surgery, or examination of the resected liver. Patients were classified according to HCC risk factors including NASH, hepatitis B and C infection, and alcohol-related. Other data collected included: demographics, ethnicity, presence of cirrhosis, tumor characteristics (size, number, vascular invasion), diabetes, hyperlipidemia, body mass index (BMI) and blood testing to calculate Model for End-Stage Liver Disease (MELD) score. Elevated AFP was defined as >20 ng/mL. Sensitivity of AFP was determined and compared between various subgroups. Results: Elevated AFP levels were detected in 64.3% of patients. AFP sensitivity was 47% for NASH-related HCC (n = 100), and 67.2% for HCC with viral or alcoholic risk factors (n = 637) (OR 0.43, 95% CI 0.28 0.66, p = 0.0001). Elevated AFP had higher sensitivity in females (71.9% vs. 61.8%, OR 1.58, 95% CI 1.1 2.27, p = 0.013), non-diabetics (67.4% vs. 57.2%, OR 0.65, 95% CI 0.47 0.89, p = 0.0093), and cirrhotics (67.1% vs. 56.8%, OR 1.55, 95% CI 1.10 2.19, p = 0.0012). AFP did not vary significantly with regard to hyperlipidemia or BMI. AFP was more sensitive in advanced disease including tumors > 5 cm, multiple tumors, or vascular invasion (all with p < 0.05). AFP did not vary with MELD score. Conclusions: Normal AFP is common in NASH-related HCC. Better tumor markers may be needed to optimally screen and diagnose NASH-related HCC. Without more effective tumor markers, HCC detection relies heavily upon imaging and liver biopsy.