Evaluation of Factors Affecting Alpha Fetoprotein Levels in Patients With Hepatocellular Carcinoma in a Large Multicenter Cohort.

BACKGROUND: Alpha-fetoprotein (AFP) is the most widely used biomarker for hepatocellular carcinoma (HCC). Given the wide variation in AFP values, conventional linear regression methods may provide an incomplete understanding of complex predictor relationships. Therefore, we utilized quantile regression to examine the association of clinical factors with AFP distribution. METHODS: In this multicenter study, we analyzed retrospective data from an adult HCC cohort, collected across nine tertiary healthcare institutions from 2003 to 2021. Quantile regression, which can model covariate effects at different specified points of the outcome distribution, was used to account for heterogeneity across the AFP value range, assessing the effects of predictors associated with AFP levels at the 0.10, 0.50, and 0.90 quantiles. Logistic regression was performed with AFP < 20 ng/mL and ≥ 20 ng/mL groups. RESULTS: The cohort included 2298 individuals with HCC, with median AFP of 13.70 ng/mL. Multivariable quantile regression determined that factors such as Asian ethnicity, elevated aspartate aminotransferase (AST), alanine aminotransferase (ALT), Barcelona Clinic Liver Cancer (BCLC) stage, hepatitis B (HBV), and hepatitis C (HCV) were associated with higher AFP, while male sex, older age, higher BMI, and elevated creatinine were associated with lower AFP levels. Compared to metabolic dysfunction-associated steatotic liver disease, HBV (β = 0.44 at Q50) and HCV (β = 0.30 at Q10; β = 0.40 at Q50) were associated with higher AFP. CONCLUSION: There is substantial heterogeneity in how clinical factors influence AFP levels across its distribution. These data may stimulate the development of more granular cut-points for AFP that differ according to disease stage and etiology.