EASL Clinical Practice Guidelines

Staging systems

Cancer classification is intended to establish prognosis and enable the selection of the adequate treatment for the best candidates. In addition, it helps researchers to exchange information and design clinical trials with comparable criteria. In patients with HCC, unlike most solid tumors, the coexistence of two life-threatening conditions such as cancer and cirrhosis complicates prognostic assessments [[99], [128]]. Thus, staging systems for this cancer should be designed with data coming from two sources. First, prognostic variables obtained from studies describing the natural history of cancer and cirrhosis. Second, treatment-dependent variables obtained from evidence-based studies providing the rationale for assigning a given therapy to patients in a given subclass.

Based on data reporting the natural history of the disease, the main clinical prognostic factors in HCC patients are related to tumor status (defined by number and size of nodules, presence of vascular invasion, extrahepatic spread), liver function (defined by Child–Pugh's class, bilirubin, albumin, portal hypertension, ascites) and general health status (defined by ECOG classification and presence of symptoms) [[129], [130], [131], [132], [133]]. Etiology has not been identified as an independent prognostic factor.

Tissue and serum biomarkers predicting prognosis have been less explored in HCC patients. Strict rules for incorporating prognostic or predictive markers into clinical practice have been published [134]. According to these rules, acceptable biomarkers should be obtained from randomized investigations, as is the case with KRAS status and response to cetuximab in colon cancer. Only in particularly compelling circumstances can prognostic or predictive markers tested in cohort studies be adopted in clinical practice. The panel recommends to incorporate biomarkers for the management of HCC when the following requirements are met: (1) demonstrate prognostic prediction in properly powered randomized studies or in training and validation sets from cohort studies; (2) demonstrate independent prognostic value in mutivariate analysis, including known clinico-pathological predictive variables; and (3) confirmation of results using the same technology in an external cohort reported by independent investigators. None of the biomarkers tested so far fulfil these criteria in HCC, but four just require external validation by independent groups: gene signatures or biomarkers from the tumor (EpCAM signature, G3-proliferation subclass, and miR-26a) [[77], [135], [136]] and adjacent tissue (poor-survival signature) [137]. Regarding serum markers, AFP levels, VEGF and Ang2 have been shown to have independent prognostic value in large cohorts of untreated advanced tumors [138]. The prognostic relevance of high AFP levels has been scarcely reported in controlled investigations [139], but has been shown to predict risk of drop-out in patients on the waiting list for liver transplantation (cut-off of 200 ng/ml, or by increases of >15 ng/ml) [[140], [141]], response to local ablation [142], response to loco-regional therapies [143] and in the outcome of advanced tumors (cut-off of 200 ng/ml [138]; 400 ng/ml [[130], [144]]). The heterogeneity of the above studies prevents the formulation of a clear recommendation, but it is advised to test levels >200 and/or >400 ng/ml as prognostic factors of poor outcome in research investigations.

Several staging systems have been proposed to provide a clinical classification of HCC. In oncology, the standard classification of cancer is based on the TNM staging. In HCC, the 7th TNM edition in accordance with the AJCC [145], which was obtained from the analysis of a series of patients undergoing resection, has several limitations [146]. First, pathological information is required to assess microvascular invasion, which is only available in patients treated by surgery (∼20%). In addition, it does not capture information regarding liver functional status or health status. One-dimensional systems, such as the Okuda staging and the Child–Pugh classification, albeit popular, serve purposes distinct to class prediction in HCC patients. Among more comprehensive staging systems, five have been broadly tested, three European (the French classification [147], the Cancer of the Liver Italian Program (CLIP) classification [130], and the Barcelona-Clínic Liver Cancer (BCLC) staging system [[148], [149]]) and two Asian (the Chinese University Prognostic Index (CUPI score) [150] and the Japan Integrated Staging (JIS), which was recently refined including biomarkers (AFP, DCP AFP-L-3) (bm-JIS) [151]). The CUPI and CLIP scores largely subclassify patients at advanced stages, with a small number of effectively treated patients. Overall, few of the most used systems or scores have been externally validated (BCLC, CUPI, CLIP, and bm-JIS), only two include the three types of prognostic variables (BCLC, CUPI) and only one assigns treatment allocation to specific prognostic subclasses (BCLC).

The current EASL–EORTC GP guidelines endorse the Barcelona-Clínic Liver Cancer (BCLC) classification for several reasons [[148], [149]]. It includes prognostic variables related to tumor status, liver function and health performance status along with treatment-dependant variables obtained from cohort studies and randomized trials. It has been externally validated in different clinical settings [[152], [153], [154]]. This is an evolving system that links tumor stage with treatment strategy in a dynamic manner enabling the incorporation of novel advancements in the understanding of the prognosis or management of HCC. In this regard, the seminal classification reported in 1999 [148] was updated with the incorporation of stage 0 (very early HCC) and chemoembolization for intermediate HCC in 2003 [99], and further modified in 2008 to incorporate sorafenib as first-line treatment option in advanced tumors [149]. As discussed below, further refinements in class stratification (for instance to incorporate biomarkers) or treatment allocation resulting from positive high-end trials are expected in the following years. The BCLC classification was first endorsed by the EASL [1], and thereafter by the AASLD guidelines for the management of HCC [56].

BCLC classification: outcome prediction and treatment allocation

The Barcelona-Clínic Liver Cancer (BCLC) classification divides HCC patients in 5 stages (0, A, B, C and D) according to pre-established prognostic variables, and allocates therapies according to treatment-related status (Fig. 3). Thus, it provides information on both prognostic prediction and treatment allocation. Prognosis prediction is defined by variables related to tumor status (size, number, vascular invasion, N1, M1), liver function (Child–Pugh's) and health status (ECOG). Treatment allocation incorporates treatment dependant variables, which have been shown to influence therapeutic outcome, such as bilirubin, portal hypertension or presence of symptoms-ECOG.

Fig. 3 Updated BCLC staging system and treatment strategy, 2011.

View Large Image | View Hi-Res Image | Download PowerPoint Slide
Early stages

Very early HCC (BCLC stage 0) is defined as the presence of a single tumor <2 cm in diameter without vascular invasion/satellites in patients with good health status (ECOG-0) and well-preserved liver function (Child–Pugh A class). Nowadays, 5–10% of patients in the West are diagnosed at this stage while in Japan the figure is almost 30% due to the widespread implementation of surveillance programs [155]. From pathological studies, though, two subclasses of tumors have been defined: vaguely nodular type – size around 12 mm without local invasiveness – and the distinctly nodular type – mean size 16 mm which might show local invasiveness. Vaguely nodular types are very well-differentiated HCCs that contain bile ducts and portal veins, have ill-defined nodular appearance and, by definition, do not have invaded structures. Distinctly nodular type show local metastases surrounding the nodule in 10% of cases, and microscopic portal invasion in up to 25% [[101], [105]]. Therefore, some tumors smaller than 2 cm are prone to locally disseminate, but others behave as carcinoma in situ and those are defined as Stage 0. Recent data have shown a 5-year survival in 80–90% of patients with resection and liver transplantation and in 70% with local ablation [[156], [157], [158], [159]]. Whether patients at very early stage can be offered local ablation as a first line treatment option is a topic of controversy. No RCT addressing this issue have been reported so far and comparison of cohort studies suffers from selection bias.

Early HCC (BCLC stage A) is defined in patients presenting single tumors >2 cm or 3 nodules <3 cm of diameter, ECOG-0 and Child–Pugh class A or B. Median survival of patients with early HCC reaches 50–70% at 5 years after resection, liver transplantation or local ablation in selected candidates [[102], [160]]. The natural outcome of these cases is ill-defined due the scarcity of reported data, but it is estimated to be a median survival of around 36 months. An improvement in survival is universal when applying the so-called treatment-dependent variables in the selection of candidates.

Tumor status is defined by size of the main nodule and multicentricity (single 2–5 cm, 3 nodules ⩽3 cm), each of these categories showing significantly different outcomes. As discussed below, single tumors beyond 5 cm are still considered for surgical resection as first option, because if modern MRI is applied in pre-operative staging, the fact that solitary large tumors remain single and with no macrovascular involvement – which might be common in HBV-related HCC – reflects a more benign biological behavior.

Variables related to liver function are relevant for candidates to resection. Absence of clinically relevant portal hypertension and normal bilirubin are key predictors of survival in patients with single tumors undergoing resection [161]. Similarly, Child–Pugh class A is the strongest prognostic variable in patients undergoing local ablation, along with tumor size and response to treatment [162]. Since liver transplantation may potentially cure both the tumor and the underlying liver disease, variables mostly related with HCC have been clearly established as prognostic factors (single tumors ⩽5 cm or 3 nodules ⩽3 cm), defining the so-called Milan criteria.

Intermediate-advanced HCC

Prognosis of HCC was assumed to be poor for unresectable cases, with a median survival of less than 1 year. Analysis of heterogeneous outcomes within 25 RCT (2 year survival 8–50%) [[131], [133], [139], [163]] leads to the identification of at least three subgroups of patients with unresectable HCC: the intermediate, advanced and end-stage classes, according to the BCLC classification.

Intermediate HCC (BCLC stage B): Untreated patients at an intermediate stage – BCLC B class (multinodular asymptomatic tumors without an invasive pattern) present a median survival of 16 months [[139], [164]], or 49% at 2 year [133]. Chemoembolization extends the survival of these patients to a median of up to 19–20 months according to RCT and meta-analysis of pooled data [139]. Nonetheless, outcome prediction is heterogeneous for BCLC B subclass patients, and has been reported to range from around 36–45 months [[165], [166], [167]] for the best responders to chemoembolization in recent series, to 11 months for the worst scenario of untreated candidates (placebo arm of the SHARP trial-BCLC B patients) [168]. A recent meta-analysis of RCT assessing outcome of patients in the control arm suggests that ascites – which contraindicates TACE treatment – is the worst prognostic factor for this subclass [133].

Advanced HCC (BCLC stage C): Patients with cancer related-symptoms (symptomatic tumors, ECOG 1–2), macrovascular invasion (either segmental or portal invasion) or extrahepatic spread (lymph node involvement or metastases) bear a dismal prognosis, with expected median survival times of 6 months [[131], [164]], or 25% at 1 year [133]. Nonetheless, it is obvious that this outcome varies according to the liver functional status and other variables. For instance, patients with preserved liver function (Child–Pugh's A class) have a median survival of 7 months [168], while those with severe liver impairment (Child–Pugh's B class) present 5 months of median life expectancy. In 2006, there was no FDA-approved first line treatment for patients with advanced HCC. This scenario has changed as a result of the data reported showing survival benefits from patients receiving sorafenib – a multi tyrosine kinase inhibitor – in advanced cases [168]. The results of this RCT represent a breakthrough in the management of HCC, as it is discussed in the molecular targeted therapies section of this document. Overall median survival in the sorafenib arm was 10.7 months, ranging from 14.7 months in BCLC B and 9.5 months in BCLC C patients.

End-stage HCC: Patients with end-stage disease are characterized by presenting with tumors leading to a very poor Performance Status (ECOG 3–4), which reflects a severe tumor-related disability. Their median survival is 3–4 months [148] or 11% at 1-year [133]. Similarly, Child–Pugh C patients with tumors beyond the transplantation threshold also have a very poor prognosis.

Concept of treatment stage migration

A proportion of patients in each stage do not fulfil all the criteria for the treatment allocation. In those cases, it is advised to offer the patient the next most suitable option within the same stage or the next prognostic stage. For instance, patients at BCLC A failing local ablation should be offered chemoembolization. Similarly, patients at BCLC B stage non-responding to chemoembolization – at least two cycles of treatment – should be offered sorafenib, as reported in the SHARP trial [[168], [169]].

Refinement of BCLC classification

Some studies challenged the capacity of BCLC to properly provide a fine stratification of patients for trial design. These studies mostly included patients at BCLC C stage of the disease [170]. The panel of experts acknowledges that the range of survival reported for patients at BCLC B (from 45 months to 11 months) and C (from 11 months to 5 months) deserves to be addressed. Further stratification of patients within each class according to liver function (Child–Pugh A versus B, or ascites), prognostic molecular biomarkers or prognostic variables (ECOG, cancer invasiveness) should be explored.

Molecular classification of HCC

Molecular classification of cancer should aid in understanding the biological subclasses and drivers of the disease and optimize benefits from molecular therapies and enrich trial populations. Few molecular classifications have been proposed in cancer. One such is the case of breast cancer, where Her2/nu status discriminates subgroups of patients with different outcome and treatment response to trastuzumab [171]. Similarly, EGFR mutational status in non-small cell lung cancer identifies a subgroup of responders to tyrosine kinase inhibitors [172]. More recently, the fact that a subgroup of patients with melanoma and BRAF mutations respond to specific B-RAF inhibitors has defined a new paradigm and subclass in the management of this cancer [173].

In HCC, no molecular subclass has been reported as responding to specific targeted therapy. Nonetheless, clear advancements in the understanding of the pathogenesis and molecular subclasses of the disease occurred during the last decade. From the biological standpoint, different tumoral classes have been characterized including a Wnt subclass, a proliferation class (with two subclasses: S1-TGF-beta and S2-EpCAM positive) and an inflammation class [[77], [137], [174], [175]]. Samples obtained from different parts of a given neoplastic nodule showed identical class stratification in 95% of cases [136]. Equally relevant, gene profiling of adjacent non-tumoral tissue defines two subgroups of patients with good and poor outcome [137]. Thus, a portrait of the field effect is currently available, although further studies are required to confirm the prognostic significance of these subclasses, and whether specific drivers within them can provide the rationale for a more stratified medicine.