Abstract
OBJECTIVE: To develop and validate a novel prognostic score combining serum total bile acid (TBA) and monocyte-to-lymphocyte ratio (MLR) for improved risk stratification in hepatocellular carcinoma (HCC) patients after radical hepatectomy. METHODS: In 508 HCC patients undergoing radical hepatectomy, we determined optimal TBA and MLR cutoffs for RFS and OS using maximally selected rank statistics. Multivariable Cox regression identified independent predictors, enabling development of a three-tiered TBA-MLR risk score (low/intermediate/high). We internally validated performance via bootstrapping (1000 iterations) and compared it against conventional biomarkers (AFP, BCLC, Child-Pugh) and inflammatory indices (SII, SIRI, NLR, PLR). Subgroup analyses assessed its ability to refine prognosis within BCLC stages and AFP categories. Concordance and overlap were assessed via Venn diagrams and Cohen's kappa coefficient. Subgroup analyses assessed the robustness of the TBA-MLR score. RESULTS: Elevated TBA (>11.7 μmol/L; HR=2.96, p<0.001) and MLR (>0.26; HR=1.64, p=0.001) independently predicted poorer RFS, while TBA (>14 μmol/L; HR=3.87, p<0.001) and MLR (>0.32; HR=1.54, p=0.036) were associated with worse OS. The TBA-MLR score stratified patients into distinct risk groups: high-risk patients had significantly lower 1/3/5-year RFS (66.9%/41.4%/19.1%) and OS (79.5%/51.1%/19.1%) versus low-risk patients (RFS:94.3%/80.8%/73.8%; OS:97.9%/90.8%/85.3%; HR=5.69 and 4.07, both p<0.001). Notably, it identified high-risk subsets within traditional low-risk categories: 22.7% of BCLC0-A patients were high-risk by TBA-MLR and had a 5-year OS of only 22.6% (vs. 86.2% in low-risk BCLC0-A patients, p<0.0001). Similarly, among patients with AFP <400 ng/mL, the high-risk group (21.9%) had a 5-year OS of 31.9% (vs. 82.7% in low-risk patients, p<0.0001). Internal validation confirmed strong predictive accuracy (C-indices: RFS 0.639, OS 0.683), with 1/3/5-year AUCs (RFS:0.657/0.660/0.771; OS:0.713/0.720/0.779) outperforming conventional biomarkers (all p<0.05). The score demonstrated minimal concordance with conventional systems (|κ|<0.06), with 16-25% of high-risk patients missed by BCLC/AFP criteria. Subgroup analyses showed consistent performance across tumor characteristics and treatments. CONCLUSION: The TBA-MLR score is a robust metabolic-immune prognostic biomarker that effectively uncovers occult high-risk biology within conventional staging systems, enabling precise postoperative risk stratification for individualized management, particularly for patients traditionally classified as low-risk (e.g., BCLC 0-A) or with non-elevated AFP.