Magnetic resonance imaging with gadoxetic acid for local tumour progression after radiofrequency ablation in patients with hepatocellular carcinoma.

Kang, Tae Wook; Rhim, Hyunchul; Lee, Jisun; Song, Kyoung Doo; Lee, Min Woo; Kim, Young-Sun; Lim, Hyo Keun; Jang, Kyung Mi; Kim, Seong Hyun; Gwak, Geum-Youn; Jung, Sin-Ho

OBJECTIVES To develop and validate a prediction model using magnetic resonance imaging (MRI) for local tumour progression (LTP) after radiofrequency ablation (RFA) in hepatocellular carcinoma (HCC) patients. METHODS Two hundred and eleven patients who had received RFA as first-line treatment for HCC were retrospectively analyzed. They had undergone gadoxetic acid-enhanced MRI before treatment, and parameters including tumour size; margins; signal intensities on T1-, T2-, and diffusion-weighted images, and hepatobiliary phase images (HBPI); intratumoral fat or tumoral capsules; and peritumoural hypointensity in the HBPI were used to develop a prediction model for LTP after treatment. This model to discriminate low-risk from high-risk LTP groups was constructed based on Cox regression analysis. RESULTS Our analyses produced the following model: 'risk score = 0.617 × tumour size + 0.965 × tumour margin + 0.867 × peritumoural hypointensity on HBPI'. This was able to predict which patients were at high risk for LTP after RFA (p < 0.001). Patients in the low-risk group had a significantly better 5-year LTP-free survival rate compared to the high-risk group (89.6 % vs. 65.1 %; hazard ratio, 3.60; p < 0.001). CONCLUSION A predictive model based on MRI before RFA could robustly identify HCC patients at high risk for LTP after treatment. KEY POINTS • Tumour size, margin, and peritumoural hypointensity on HBPI were risk factors for LTP. • The risk score model can predict which patients are at high risk for LTP. • This prediction model could be helpful for risk stratification of HCC patients.