Original Research Open Access Gastrointestinal Imaging Hanyu Jiang *, Chongtu Yang *, , Yidi Chen , Yanshu Wang , Yuanan Wu , Weixia Chen , Maxime Ronot , Victoria Chernyak , Kathryn J. Fowler Mustafa R. Bashir **, Bin Song ** Figure 1: Study inclusion and exclusion flowchart. BCLC = Barcelona Clinic Liver Cancer, HCC = hepatocellular carcinoma. Figure 2: Illustration of the advanced-stage recurrence after resection (ASRAR) scoring system. Figure 3: Preoperative axial extracellular contrast agent-enhanced (A–E) MRI and (F–H) follow-up CT images in a 71-year-old male patient with chronic hepatitis B and a serum neutrophil count of 4.2 × 10 9/L. A 5.6-cm mass with a 1.9-cm satellite nodule was detected in segments V and VIII. (A) The mass (asterisks, A–E) shows mild-to-moderate hyperintensity on T2-weighted images, (B) hypointensity on T1-weighted noncontrast-enhanced images, (C) less than 50% hyperenhancement on late arterial phase images, (D) nonperipheral washout and incomplete enhancing capsule on portal venous phase images, and (E) marked diffusion restriction on diffusion-weighted images ( b = 800 sec/mm 2). The mass was confirmed at histopathologic analysis as moderately differentiated hepatocellular carcinoma with microvascular invasion. With an advanced-stage recurrence after resection score of 19.8 points (4.2 + 5.6 + 10), this patient was assigned to the high-risk group (≥15 points) for advanced-stage recurrence. At follow-up, (F) noncontrast-enhanced, (G) arterial phase, and (H) portal venous phase CT images acquired at day 70 after surgery show multifocal intrahepatic recurrence and tumor thrombus in the left portal vein (black arrow, H). Figure 4: Graphs show discriminative performances. The concordance index in the (A) training and (B) test sets, and time-dependent areas under the receiver operating characteristic curve in the (C) training and (D) test sets measured model discriminatory performance at different points. Bootstrapping analysis (1000 times) was performed to correct for optimism of the advanced-stage recurrence after resection (ASRAR) score concordance index in the training set (A). Graphs show the calibration and decision curves of the ASRAR score and currently used major staging systems for time to advanced-stage recurrence. The Brier score evaluates model calibration in the (E) training and (F) test sets. Decision curves are plotted for time to advanced-stage recurrence at 2 years in the (G) training and (H) test sets. AJCC = American Joint Committee on Cancer, BCLC = Barcelona Clinic Liver Cancer, CNLC = China Liver Cancer, HKLC = Hong Kong Liver Cancer, ITA.LI.CA = Italian Liver Cancer, UHPI = unresectable hepatocellular carcinoma prognostic index, UICC = Union for International Cancer Control. Figure 5: Graphs show survival outcomes in patients at high risk (≥15 points) and low risk (<15 points) of recurrence for time to advanced-stage recurrence in the (A) training and (B) test sets, and all-stage recurrence-free survival in the (C) training and (D) test sets. Graphs show advanced-stage recurrence-free survival in the (E) training and (F) test sets and overall survival in the (G) training and (H) test sets for patients without adjuvant therapies. Figure 6: Graphs show all-stage recurrence-free survival outcomes in patients with and without adjuvant therapies for (A) the entire propensity score-matched cohort ( P = .31), (B) the high-risk group ( P = .02), and (C) the low-risk group ( P = .67).