Preoperative Remnant Liver Function Evaluation Using a Routine Clinical Dynamic Gd-EOB-DTPA-Enhanced MRI Protocol in Patients with Hepatocellular Carcinoma.

BACKGROUND To investigate the clinical feasibility of preoperative routine clinical dynamic Gd-EOB-DTPA-enhanced MRI alone to predict post-hepatectomy liver failure (PHLF) in patients with hepatocellular carcinoma (HCC). METHODS 116 patients with HCC who underwent liver resection in Southwest Hospital from 2014 through 2017 were selected in this retrospective cohort study. The remnant function (RF) of the liver RFUR and RFRE15 were calculated by the sum of the uptake rate (UR) or relative enhancement at 15 min (RE15) from dynamic Gd-EOB-DTPA-enhanced MR images in the remnant liver regions, and standardized by standard liver volume (SLV) to generate sRFUR (standardized RFUR) and sRFRE15 (standardized RFRE15). Student's t test or Mann-Whitney U test, logistic regression, and ROC analyses were used to test the associations of preoperative RFUR, sRFUR, RFRE15, sRFRE15, the remnant liver volume (RLV)/SLV, ICG retention rate at 15 min (ICG R15) and sRFICG-K [ICG clearance rate (ICG-K) × RLV/SLV] with PHLF. RESULTS 28 patients were found to have PHLF, who showed lower RFUR, sRFUR, RFRE15, sRFRE15, RLV/SLV, sRFICG-K, and higher ICG R15 than patients without PHLF (p < 0.001 for all). After adjusting for clinical parameters, RFUR (p = 0.001), sRFUR (p = 0.001), RFRE15 (p = 0.002), or sRFRE15 (p = 0.003) was found to be independently significant indicator in multivariable logistic regression, respectively. RFUR (0.882) and sRFUR (0.882) had larger AUCs than RLV/SLV (0.731, p = 0.008; p = 0.005), ICG R15 (0.765, p = 0.039; p = 0.044) and sRFICG-K (0.767, p = 0.031; p = 0.023). RFRE15 (0.845) and sRFRE15 (0.839) had larger AUCs than RLV/SLV (0.731, p = 0.027; p = 0.025). CONCLUSIONS The remnant liver function parameters preoperatively estimated from a routine clinical dynamic Gd-EOB-DTPA-enhanced MRI protocol can predict PHLF in patients with HCC, and may be better predictors than conventional methods.