Evaluating AFP–liver reserve discordance and microvascular invasion in resected and transplanted hepatocellular carcinoma.

Abstract

e16262 Background: The prognosis of hepatocellular carcinoma (HCC) is influenced by tumor biology and liver function. Mismatch between tumor aggressiveness and liver function can obscure the accuracy of current risk stratification models especially when tumors occur in patients with preserved liver function. Alpha-fetoprotein (AFP) reflects aggressive tumor biology through association with larger tumor size, vascular invasion, and higher grading. MELD score reflects liver function and reserve in HCC patients by quantifying the liver dysfunction through various laboratory and liver function parameters. Despite the recognized importance of tumor characteristics and liver function, existing HCC staging systems have not defined the combined prognostic significance using AFP and MELD score. Methods: Among 366 patients (199 transplant, 167 resection), we evaluated whether preoperative AFP in relation to liver reserve was associated with microvascular invasion (MVI). Multivariable logistic regression included 276 patients (90 excluded due to missing data), adjusting for age, tumor diameter, multifocality, and locoregional therapy. Patients were stratified into four discordance phenotypes based on AFP ( > 40 ng/mL) and liver reserve (above or below median MELD). Interaction terms evaluated effect modification by surgery type. Secondary analyses addressed capsular/lymphatic invasion and tumor differentiation (proportional odds ordinal regression). Statistical analyses were performed in Python (statsmodels) with significance set at a two-sided p < 0.05. Results: Larger tumor diameter was independently associated with higher odds of MVI (OR 1.14 per cm; 95% CI 1.04–1.25; p = 0.005). Patients with a high-AFP/low-reserve phenotype showed higher odds of MVI (OR 1.96), though not statistically significant (p = 0.378), warranting further validation. There was no significant interaction between discordance phenotype and surgery type (all p > 0.30), indicating that the association between discordance phenotypes and MVI was similar in both resection and transplant cohorts. Surgery type showed a trend toward lower odds of MVI in transplant recipients (OR 0.41; p = 0.089). Conclusions: AFP–liver reserve discordance may refine risk stratification for microvascular invasion in HCC, as the association between these phenotypes and MVI risk was not significantly modified by surgery type. These findings support incorporating AFP-liver reserve phenotypes into a unified staging or risk framework to enhance prognostic accuracy in HCC beyond current AFP or MELD based models.