In addition, progression due to NEH (2.42, 1.32-4.44; or EHG 2.39, 1.15-4.96) was added as independent OS predictors in patients with radiologic tumor progression (Table 2). We excluded 23/147 patients from the analysis of PPS because they did not have at least one image evaluation and those 39 who had not presented radiologic

progression at the time of database lock. Median PPS in the 85 patients with radiologic progression was 9.85 months (95% CI: 7.3-12.5). BCLC stage, PS, and Child-Pugh status, which were learn more evaluated at the time of progression, together with progression due to NEH were the independent predictors of PPS (Table 3). The PPS of the previously defined subgroup of patients who would still be fit for second-line treatment was 13.6 months (95% CI: 9-18.2)

(Fig. 3). PPS was significantly different (P = 0.034) according to BCLC stage at progression and according to progression pattern (P = 0.013) (Figs. A2 and A3 in Supporting Material). Thereby, BCLC-C patients with NEH had a significantly worse PPS than those without it (7.1 versus 14.9 months, P = 0.02) (Fig. 4). Systematic review studies in lung,[6, 7] breast,[8] and colorectal[15] cancer have stressed the need to analyze PPS as a potential confounder for OS. Interestingly, no study has established the correlation between progression and survival in patients with HCC, and there are no data about GDC-0449 the predictors of survival after progression. Furthermore, no investigation has focused on the potential outcome differences according to the pattern of progression. As a whole, the current use of TTP as a signal to detect therapeutic efficacy is not supported by robust data gathered using proper statistical

methods that take into account time-dependent covariates. Our results show for the first time that tumor progression at imaging has a significant correlation with OS in patients with HCC and, thus, validate the use of TTP as a valid endpoint in early phase studies to evaluate the potential efficacy of novel molecular agents. Together with this association, we show that survival after progression (PPS) is significantly different according to the progression patterns. Indeed, PPS may correlate better with OS than PFS.[5, 7, 8, 15] The review of www.clinicaltrials.gov and recently MCE published trials in breast, lung, colorectal, and HCC shows that progression pattern is not considered in the evaluation of the patients to define prognosis and/or to stratify patients prior to randomization. Interestingly, a panel of several leading experts in oncology has stressed the need to further dissect the prognostic meaning of the different types of progression that may be encountered and has called for prospective studies to characterize PPS and its outcome predictors,[5] as we have done in our population of HCC patients.