MRECIST Assessment For Patients With Unresectable Hepatocellular Carcinoma Receiving TACE And Sorafenib Therapy

Background and objectivesTransarterial chemoembolization(TACE) and Sorafenib are currently standard treatments for intermediate and advanced-stage HCC, respectively. The modified Response Evaluation Criteria in Solid Tumors(m RECIST) plays a pivotal role in the treatment response evaluation. And several studies have stated that m RECIST response is an independent predictor for OS after TACE and/or Sorafenib therapy. However, the m RECIST criteria had its limitations: for irregular shape of tumors, the consistency between the observers is not very good. Furthermore, several studies have confirmed that hand-foot-skin reaction(HFSR) could be used as an early significant prognostic factor of survival. Our first analysis addressed the question: Can combination of HFSR and m RECIST predict the survival more comprehensively than each in patients with HCC receiving TACE combined with Sorafenib therapy?Secondly, there has been a controversy regarding the best time point to evaluate treatment responses can accurately predict the survival. We found the likely positive correlations between the tumor burden in terms of tumor size and tumor number and the time point response assessment in the published studies: for patients with higher tumor burdens, that means the larger in size and the more in number of lesions, the optimal time points to assess tumor response were at a later post-treatment phase. Therefore, our second analysis sought to assess the prognostic value of initial response and best response when the patients were stratified according to the tumor load.Methods 1. 176 consecutive intermediate-advanced HCC patients treated with combination therapy between January 2009 and December 2013 were enrolled. The therapeutic responses were assessed by m RECIST and HFSR criteria at 1, 2 and 3 months, respectively. Uni/multivariate Cox regressions were used to investigate the earliest time when treatment responses could be accurately assessed. Then according to the m RECIST and HFSR assessed at that time, SMART(Sorafenib with Modified RECIST Assessment plus hand-foot-skin Reaction in TACE) prognostic evaluation was developed: SMART A, responders on both assessments; SMART B, responders on either of assessment and SMART C, non-responders on both assessments. Next, we compared the prognostic value of SMART in comparison with m RECIST or HFSR. Furthermore, we investigated whether it can predict the OS in several clinical subgroups. 2. A total of 350 patients treated with TACE between January 2010 and December 2014 were enrolled in our study. Treatment responses were assessed using m RECIST criteria. Spline-based analysis was used to define the best cutoff value of tumor size and number. Kaplan–Meier and Cox regression analyses were used to explore differences in OS between responders and non-responders, defined by initial m RECIST response and best m RECIST response.Results 1. The earliest time at which the responses of m RECIST and HFSR correlated with the survival was 2 months after therapy. The SMART stratified patients with three different prognosis; the SMART A had the longest median overall survival, followed by SMART B and SMART C(30.5, 17.4, and 8.3 months, respectively; P<0.001).Compared with m RECIST and HFSR, the SMART had the highest likelihood ratio and C-index, and lowest Akaike information criterion(AIC), demonstrating that the SMART had a better performance in predicting the survival. Moreover, the SMART can predict the survival when the patients were stratified according to BCLC stage, ECOG score and AFP value. 2. Spline-based analysis revealed that the best cutoff value of tumor size was 8 cm and that of 2 for tumor number. On this basis, tumor load was stratified into two groups: a total of 184 patients with 8 cm or less in size and 2 or less in number were assigned to low-tumor load group; Another 166 patients were assigned to high-tumor load group. In low-tumor load group, differences in median survival between responders and non-responders were statistically significant for both initial response(37.8 versus 20.8 months, P < 0.001) and the best response(34.6 versus 11.9 months, P < 0.001). Multivariable Cox regression analysis showed that the two are both independent predictors of survival in low-tumor load group. In high-tumor load group, no significant statistic difference in median survival was observed between responders and non-responders for initial response(16.9 versus 9.2 months, P=0.242); whereas there was significant statistic difference in median survival between responders and non-responders for best response(20.8 versus 8.3 months, P= 0.003). Multivariable Cox regression analysis showed that the best response, rather than initial response, was an independent predictor of survival.Conclusions 1. The SMART prognostic evaluation, based on combining the m RECIST and HFSR criteria, is clinical useful to predict survival of HCC patients at an early time point after combination therapy. 2. The response evaluation should be further stratified by the tumor load. For patients with low tumor load, the initial response and the best response correlate well with the survival. Whereas, for patients with high tumor load, the best response, rather than the initial response, correlates better with the survival.