A simplified MRI‑based risk score and nomogram to predict initial objective response in unresectable hepatocellular
Yi-Kang Wang1, Hua-Guo Feng2, Yan-Han Liu3
1Second Affiliated Hospital of Chongqing Medical University, Chongqing, China.
Purpose:
This study aimed to develop and validate a simplified risk score based on pretreatment Gd‑EOB‑DTPA MRI features for the noninvasive prediction of initial objective response rate (ORR) in patients with unresectable hepatocellular carcinoma (uHCC) undergoing triple therapy (hepatic arterial infusion chemotherapy [HAIC], tyrosine kinase inhibitors, and immune checkpoint inhibitors).
Methods:
This retrospective study enrolled 223 uHCC patients from two medical centers who underwent Gd‑EOB‑DTPA‑enhanced MRI prior to initiating triple therapy (training set: n = 127; external test set: n = 96). Imaging features were evaluated and extracted. Independent predictors of initial ORR were identified by multivariate logistic regression to build a simplified imaging risk score (I-score), and a nomogram was constructed to visualize the final predictive model. Model performance was assessed by AUC, calibration (Hosmer‑Lemeshow test), confusion matrix, and decision curve analysis (DCA), with internal bootstrap and external validation. SHAP analysis provided interpretability. Patients were stratified into low‑ and high‑risk groups, and Kaplan‑Meier curves for progression‑free survival (PFS) and overall survival (OS) were compared.
Results:
Three MRI features (irregular morphology, rim enhancement, peritumoral low signal) were independent negative predictors of initial ORR. The simplified I-score was formulated as: 1×(irregular morphology) + 1 × (rim enhancement) + 2 × (peritumoral low signal), with an optimal cut-off of 1.5. Following LASSO regression, BMI > 22.55, PA < 182.5 mg/L, NLR > 2.37, and I-score ≥ 2 were included in the nomogram. The nomogram achieved AUCs of 0.890 in training and 0.841 in validation for ORR prediction, demonstrating good calibration (P > 0.05) and clinical benefit. SHAP analysis identified I-score as the most important contributor. The model significantly outperformed BCLC, ALBI, Child‑Pugh, and Six‑and‑Twelve staging systems (all P < 0.05). High‑risk patients had significantly worse PFS and OS than low‑risk patients.
Conclusion:
The MRI‑based simplified risk score and the derived nomogram offer a noninvasive and precise method for predicting initial ORR and classifying prognosis in uHCC patients undergoing triple therapy.
