Serum Protein Panel Facilitates Surgical Decision-Making for Barcelona Stage B Hepatocellular Carcinoma

To develop a serum proteomic-based prognostic model for predicting surgical outcomes in BCLC stage B hepatocellular carcinoma (HCC) patients, addressing the clinical heterogeneity that limits uniform treatment recommendations.

Preoperative serum samples from HCC patients undergoing curative resection were retrospectively collected and subjected to proteomic profiling using nanoparticle-enhanced data-independent acquisition mass spectrometry (DIA-MS). A prognostic risk model was developed in a training cohort (N = 52, BCLC stage B) using 5-fold cross-validated Lasso-Cox machine learning, with overall survival (OS) as the primary endpoint. The model was validated in an independent BCLC stage B cohort (N = 22), extended to a single huge HCC cohort (N = 26, BCLC stage A HCC with a solitary tumor measuring at least 10 cm), and benchmarked against a comparison cohort (N = 190, BCLC 0/A without huge HCC). Orthogonal validation was performed by parallel reaction monitoring (PRM) in an exploration cohort (N = 51, BCLC stage B HCC).

Among 2,381 identified proteins, a three-protein panel (LCP1, DOK3, CNN2) was selected to construct the model. The model demonstrated strong predictive accuracy for 3-year OS in the training cohort (AUC = 0.936, sensitivity = 0.988, specificity = 0.834), outperforming the clinical Kinki criteria, in the validation cohort (AUC = 0.812, sensitivity = 0.875, specificity = 0.786) and in the extended cohort of huge HCC patients (AUC = 0.758, sensitivity = 0.709, specificity = 0.852). Notably, low-risk BCLC stage B patients showed comparable prognosis to the comparison cohort of BCLC 0/A patients, suggesting potential surgical benefit. The PRM validation in an independent exploration cohort confirmed the model's prognostic value (AUC = 0.800, sensitivity = 0.870, specificity = 0.700).

This serum proteomic-based prognostic model accurately predicts surgical outcomes for BCLC stage B HCC patients and outperforms clinical stratification. By identifying low-risk patients who may benefit from surgery, this tool facilitates personalized treatment planning and has the potential to improve survival and quality of life.