Risk stratification for early recurrence after resection in patients with intermediate stage hepatocellular carcinoma

초록

Background and aims: It is unclear which patients will benefit from resection at intermediate stage of hepatocellular carcinoma (HCC). We aimed to identify high-risk patients for early recurrence in patients resectable for intermediate-stage HCC. Method: This multicenter, retrospective study involved 1,686 patients who underwent resection or transarterial chemoembolization (TACE) for intermediate-stage HCC (2008–2019). Multivariable Cox proportional analysis for identifying high-risk patients treated with resection was performed. A prediction model for 2-year recurrence-free survival (RFS) was developed in the training cohort and validated in the validation cohort. 2-year RFS in each risk group was compared to those treated with TACE after propensity-score matching. Results: During median follow-up of 31.4 months, 2-year RFS was significantly higher in the resection group (28.5%, n = 480) than in the TACE group (71.5% n = 1,206) (adjusted hazard ratio [aHR] = 1.471, 95% CI = 1.199–1.803, P < 0.001). Higher alpha-fetoprotein (aHR = 0.202), ALBI grade (aHR = 0.709), tumor number (aHR = 0.404), and maximal tumor size (aHR = 0.323) were significant risk factors for 2-year RFS in patients with resection. The newly developed Surgery Risk score in BCLC-B (SR-B score) with four variables showed an area under the curve of 0.801 for 2-year RFS and was externally validated. Based on risk stratification by the SR-B score, low-risk patients had a significantly higher 2-year RFS (training: aHR = 5.834; validation: aHR = 5.675) than high-risk patients (all P < 0.001). In a propensityscore matched cohort, low-risk patients treated with resection had a significantly higher 2-year RFS than those with TACE (aHR = 3.891); high-risk patients had a comparable 2-year RFS than those with TACE (aHR = 0.816). Conclusion: Resection may be beneficial to resectable patients with intermediate-stage HCC based on the SR-B score.