Development of a Model including MRI Features for Predicting Advanced-stage Recurrence of Hepatocellular Carcinoma after Liver Resection
Hanyu Jiang,Chongtu Yang,Yidi Chen,Yanshu Wang,Yuanan Wu,Weixia Chen,Maxime Ronot,Victoria Chernyak,Kathryn J Fowler,Mustafa R Bashir,Bin Song
DOI: https://doi.org/10.1148/radiol.230527
IF: 19.7
Radiology
Abstract:Background Identifying patients at high risk for advanced-stage hepatocellular carcinoma (HCC) recurrence after liver resection may improve patient survival. Purpose To develop a model including MRI features for predicting postoperative advanced-stage HCC recurrence. Materials and Methods This single-center, retrospective study includes consecutive adult patients who underwent preoperative contrast-enhanced MRI and curative-intent resection for early- to intermediate-stage HCC (from December 2011 to April 2021). Three radiologists evaluated 52 qualitative features on MRI scans. In the training set, Fine-Gray proportional subdistribution hazard analysis was performed to identify clinical, laboratory, imaging, pathologic, and surgical variables to include in the predictive model. In the test set, the concordance index (C-index) was computed to compare the developed model with current staging systems. The Kaplan-Meier survival curves were compared using the log-rank test. Results The study included 532 patients (median age, 54 years; IQR, 46-62 years; 465 male patients), 302 patients from the training set (median age, 54 years; IQR, 46-63 years; 265 male patients), and 128 patients from the test set (median age, 53 years; IQR, 46-63 years; 108 male patients). Advanced-stage recurrence was observed in 38 of 302 (12.6%) and 15 of 128 (11.7%) of patients from the training and test sets, respectively. Serum neutrophil count (109/L), tumor size (in centimeters), and arterial phase hyperenhancement proportion on MRI scans were associated with advanced-stage recurrence (subdistribution hazard ratio range, 1.16-3.83; 95% CI: 1.02, 7.52; P value range, <.001 to .02) and included in the predictive model. The model showed better test set prediction for advanced-stage recurrence than four staging systems (2-year C-indexes, 0.82 [95% CI: 0.74, 0.91] vs 0.63-0.68 [95% CI: 0.52, 0.82]; P value range, .001-.03). Patients at high risk for HCC recurrence (model score, ≥15 points) showed increased advanced-stage recurrence and worse all-stage recurrence-free survival (RFS), advanced-stage RFS, and overall survival than patients at low risk for HCC recurrence (P value range, <.001 to .02). Conclusion A model combining serum neutrophil count, tumor size, and arterial phase hyperenhancement proportion predicted advanced-stage HCC recurrence better than current staging systems and may identify patients at high risk. Published under a CC BY 4.0 license. Supplemental material is available for this article. See also the editorial by Tsai and Mellnick in this issue.