A Dynamic Online Nomogram Based on Gd-EOB-DTPA-Enhanced MRI and Inflammatory Biomarkers for Preoperative Prediction of Pathological Grade and Stratification in Solitary Hepatocellular Carcinoma: A Multicenter Study

Background

Hepatocellular carcinoma (HCC) is an inflammatory cancer. We aimed to explore whether preoperative inflammation biomarkers compared to the gadoxetic acid disodium (Gd-EOB-DTPA) enhanced MRI can add complementary value for predicting HCC pathological grade, and to develop a dynamic nomogram to predict solitary HCC pathological grade.

Methods

331 patients from the Institution A were divided chronologically into the training cohort (n = 231) and internal validation cohort (n = 100), and recurrence-free survival (RFS) was determined to follow up after surgery. 79 patients from the Institution B served as the external validation cohort. Overall, 410 patients were analyzed as the complete dataset cohort. Least absolute shrinkage and selection operator (LASSO) and multivariate Logistic regression were used to gradually filter features for model construction. The area under the receiver operating characteristic curve (AUC) and decision curve analysis were used to evaluate model's performance.

Results

Five models of the inflammation, imaging, inflammation+AFP, inflammation+imaging and nomogram were developed. Adding inflammation to imaging model can improve the AUC in training cohort (from 0.802 to 0.869), internal validation cohort (0.827 to 0.870), external validation cohort (0.740 to 0.802) and complete dataset cohort (0.739 to 0.788), and obtain more net benefit. The nomogram had excellent performance for predicting high-grade HCC in four cohorts (AUCs: 0.882 vs. 0.869 vs. 0.829 vs. 0.806) with a good calibration, and accessed at https://predict-solitaryhccgrade.shinyapps.io/DynNomapp/. Additionally, the nomogram obtained an AUC of 0.863 (95% CI 0.797–0.913) for predicting high-grade HCC in the HCC≤ 3 cm. Kaplan–Meier survival curves demonstrated that the nomogram owned excellent stratification for HCC grade (P < 0.0001).

Conclusion

This easy-to-use dynamic online nomogram hold promise for use as a noninvasive tool in prediction HCC grade with high accuracy and robustness.