Liver imaging reporting and data system category 4 observations in MRI: Risk factors predicting upgrade to category 5.

Abstract

To identify demographic and imaging features in magnetic resonance imaging (MRI) that are associated with upgrade of Liver Imaging Reporting and Data System (LI-RADS) category 4 (LR-4) observations to category 5 (LR-5), and to assess their effects on risk of upgrade and time to upgrade. Institutional Review Board approval was obtained for this retrospective, dual-institution Health Insurance Portability and Accountability Act (HIPAA)-compliant study. Radiologists reviewed 1.5T and 3T MRI examinations for 181 LR-4 observations in 139 patients, as well as follow-up computed tomography (CT) and MRI examinations and treatment. A stepwise multivariate Cox proportional hazards model analysis was performed to identify predictive risk factors for upgrade to LR-5, including patient demographics and LI-RADS imaging features. Overall cumulative risk of upgrade was calculated by using the Kaplan-Meier method. The cumulative risks were compared in the presence/absence of significant predictive risk factors using the log-rank test. The independent significant predictive risk factors in the 56 LR-4 observations that upgraded to LR-5 were mild-moderate T2 hyperintensity (P < 0.001; hazard ratio = 1.84), growth (P < 0.001; hazard ratio = 3.71), and hepatitis C infection (P = 0.02; hazard ratio = 1.69). The overall 6-month cumulative risk of upgrade was 32.7%. The 6-month cumulative risk rate was significantly higher in the presence of T2 hyperintensity (P = 0.03; 48.1% vs. 25.4%). For LR-4 observations, mild-moderate T2 hyperintensity, threshold growth, and hepatitis C infection are associated with significantly higher risk of upgrade to LR-5. Although mild-moderate T2 hyperintensity was the most useful risk factor for predicting upgrade, actual risk level was only mildly elevated, and the risk of upgrade associated with LR-4 observations is similar across subtypes. 3 Technical Efficacy: Stage 5 J. MAGN. RESON. IMAGING 2017;46:783-792.