Prognostic prediction in HCC: Did anybody expect it to be easy?

Staging and prognostication of patients with hepatocellular carcinoma (HCC) is a very controversial area where a wide consensus is lacking.1 A well-validated and internationally accepted staging system is critical in order to fulfill several needs.2 Firstly, it should allow healthcare providers to give accurate information to patients about long-term life expectancy. In addition, it should stratify patients into subgroups with significantly different outcomes. Obviously, the outcome of each strata should be reproduced in different studies applying the same system, as one of the aims of any classification is to facilitate comparison between studies. Finally, a staging system should help healthcare providers to indicate the most adequate treatment and to predict the outcome after therapy.2 Accordingly, the task is not simple, and as will be discussed below, it is far from being solved. Part of the difficulty is related to the fact that more than 90% of HCC cases appear in a cirrhotic liver.3 Thus, prognostication should consider both the tumor extent and the degree of liver function impairment.3 This is different from most solid tumors in which the outcome is predicted by the tumor itself and by the feasibility of applying effective treatment. The article by Henderson et al.4 summarizes the effort conducted by a Consensus Panel set up by the AHPBA to review the available systems to stage HCC patients and suggests which of the several prognostic models should currently be used. The proposals that were discussed included the Okuda staging system,5 the Cancer of the Liver Italian Program (CLIP),6 the BCLC,7 the CUPI,8 the modified TNM,9, 10 and the Japanese proposal recommended by the IHPBA and supported by the UICC (Makuuchi et al., manuscript submitted). Among the clinical methods, the panel endorsed the CLIP score, because it has been partially validated and is the easiest to use. In addition, they recommended the modified TNM as the best pathologic system. The effort made by the panel has to be acknowledged, but unfortunately it seems they have stretched themselves too far by producing a statement in support of one of the systems, as they simultaneously accept that no system fulfils the requirements of acceptance because of several limitations. These limitations are listed in the manuscript and unequivocally argue against the panel's recommendation. The use of scoring systems that divide patients into strata with different prognoses, without any link to treatment indication or the ability to predict the outcome after therapy, has no major clinical value. Furthermore, despite some positive suggestions, it is unfortunate to note that even if some systems such as the CLIP5 or CUPI7 properly stratify patients according to outcome, the survival of the strata has not been reproduced in studies from different geographic origins,11 nor do they retain their prognostic power in comparison with other staging systems.7 This lack of reproducibility suggests that the characteristics and evolutionary stages of the patients recruited in the investigations are heterogeneous and/or that some important predictors that may determine a different outcome have not been taken into account. Thereby, most systems are useful for identification of end-stage patients, but lack the accuracy needed to identify early stage patients and distinguish them from intermediate cases (the survival of the best CLIP strata is 50% at 3 years). At the same time, pathologic systems have proven to be useful for resected HCC, but their usefulness after transplantation is controversial, and can not be applied to non-surgical patients. Furthermore, early stage patients are usually treated by surgery or percutaneous ablation (the reason for no treatment represents an important bias) and thus, prognostication at this stage should always incorporate treatment-dependent variables. With all these comments in mind, it might finally appear more useful to postpone the search for a single staging system to serve all HCC patients. Instead, efforts should be focused on the development of prognostic models for any relevant evolutionary stage of the disease (early, advanced, and terminal), each of them with specific needs.3, 12 Hence, with the current evidence, the most reliable statement that the Consensus Panel has released is that neither of the available staging systems are free of limitations and in the future other factors such as etiology, treatment variables, and new markers may have to be incorporated. In that sense, it is expected that molecular markers may overcome some of the limitations currently faced when using merely clinical or pathological variables. Accordingly, no answer can be given to the question of which system should be used to stage HCC patients. Thus, instead of supporting the use of the easiest system at the cost of reduced clinical usefulness and reliability, the request should be to further develop prognostic investigations. If designed to answer well-defined questions, they will very likely deal with specific scenarios (transplantation, resection, non-surgical HCC) and again skip the search for a common model for all patients.