Heterogeneous Type 4 Enhancement of Hepatocellular Carcinoma on Dynamic CT Is Associated With Tumor Recurrence After Radiofrequency Ablation

OBJECTIVE. The aim of this study was to predict recurrence of hepatocellular carcinoma (HCC) from baseline dynamic CT images. MATERIALS AND METHODS. This retrospective study included 191 consecutive patients who underwent surgical resection or radiofrequency ablation (RFA) between January 2005 and September 2009 for the treatment of HCC. Enhancement on pretreatment arterial and portal phase dynamic CT images was classified into one of the four following enhancement patterns: Types 1 and 2 are homogeneous enhancement patterns without or with increased arterial blood flow, respectively; type 3 is a heterogeneous enhancement pattern with septations; and type 4 is an irregularly shaped ring structure enhancement pattern. Predictive factors for tumor recurrence including dynamic CT enhancement pattern were also evaluated. Moreover, risk factors including recurrence type (i.e., tumor number >= 10, portal vein invasion, or both) were evaluated in RFA-treated cases. RESULTS. Among 60 patients who underwent surgical resection, no statistical association was observed between dynamic CT enhancement patterns and recurrence rate. In contrast, in the 131 patients who underwent RFA, cumulative recurrence rates for each enhancement pattern were significantly different: Recurrence rates 2 years after RFA for patients with type 1, 2, 3, and 4 were 26.6%, 46.9%, 38.6%, and 77.8%, respectively (p = 0.042). Recurrence, which was defined as the presence of 10 or more nodules, portal vein invasion, or both occurred in nine of 131 patients (6.9%) in the RFA group. A multivariate Cox proportional hazards analysis revealed that the type 4 dynamic CT enhancement pattern is an independent factor for HCC recurrence (hazard ratio, 27.68; 95% CI, 6.82-112.33; p < 0.001). CONCLUSION. The pretreatment type 4 dynamic CT enhancement pattern can potentially be used to predict recurrence of HCC after RFA treatment.