Imperial college London-MRes in Translational Medicine (2010-2011)

Validation of the ALBI grade in HCC

The ALBI grade provides objective hepatic reserve phenotyping

across each BCLC stage of hepatocellular carcinoma.

David J. Pinato*1, Rohini Sharma*1, Elias Allara2, Clarence Yen1, Tadaaki Arizumi3, Keiichi Kubota4, Dominik Bettinger5, Jeong Won Jang6, Carlo Smirne7, Young Woon Kim6, Masatoshi Kudo3, Jessica Howell1, Ramya Ramaswami1, Michela E. Burlone7, Vito Guerra8, Robert Thimme5, Mitsuru Ishizuka4, Justin Stebbing1, Mario Pirisi7, Brian I. Carr9.

1. Department of Surgery & Cancer, Imperial College London, Hammersmith Hospital, Du Cane Road, W120HS London, UK.

2. Independent Biostatistician, Cambridge, UK.

3. Department of Gastroenterology and Hepatology, Kinki University School of Medicine, Osaka-Sayama, Osaka, Japan

4. Department of Gastroenterological Surgery, Dokkyo Medical University, Mibu, Tochigi 321-0293, Japan

5. Department of Medicine II, University Hospital Freiburg, Freiburg, Germany.

6. Department of Internal Medicine, The Catholic University of Korea Incheon St. Mary’s Hospital, Seoul, Republic of Korea.

7. Department of Translational Medicine, Università degli Studi del Piemonte Orientale “A. Avogadro”, Via Solaroli 17, Novara, Italy.

8. IRCCS De Bellis, National Institute for Digestive Diseases, Castellana Grotte, Italy

9. Izmir Biomedicine and Genome Center, Dokuz Eylul University, Izmir, Turkey

To whom correspondence should be addressed:

Dr David James Pinato, MD MRes MRCP (UK) PhD

NIHR Academic Clinical Lecturer in Medical Oncology, Imperial College London Hammersmith Campus, Du Cane Road, W12 0HS, London (UK)

E mail: david.pinato@imperial.ac.uk

Keywords: Prognosis, hepatocellular carcinoma, liver failure, ALBI, albumin, bilirubin.

Word Count: 3980

Tables: 4 Figures: 2

Running Head: Validation of the ALBI grade in HCC.

*Indicates joint-first authorship.

Authors Contribution:

Study concept and design: DJP, RS, BIC.

Acquisition of data: DJP, TA, JWJ, CS, YWK, MK, MP, DB, RT, CY, DB.

Analysis and interpretation of data: DJP, EA RR, MP, RS, JS, JH.

Drafting of the manuscript: DJP, RS, EA, BIC, JH.

Critical revision of the manuscript for important intellectual content: All the authors.

Statistical analysis: DJP, RR, EA, JH.

Obtained funding: DJP, JWJ, RS, MP.

Administrative, technical, or material support: MP, MK, JWJ, RT, JS.

Study supervision: DJP, JWJ, MK, MP, RS, JS.

Funding:

No specific funding was obtained to support the conduction of this study.

Conflict of Interest:

The authors have no conflicts of interest to disclose.

Lay Summary:

Liver failure is a key determinant influencing the natural history of hepatocellular carcinoma (HCC). In this large multicenter study we externally validate a novel biomarker of liver functional reserve, the ALBI grade, across all the stages of HCC.

Abstract.

Background & Aims: Overall survival (OS) is a composite clinical endpoint in hepatocellular carcinoma due to the mutual influence of cirrhosis and active malignancy in dictating patient’s mortality. The ALBI grade is a recently described index of liver dysfunction in HCC, based solely on albumin and bilirubin levels. Whilst accurate, this score lacks cross-validation, especially in intermediate-stage HCC, where OS is highly heterogeneous.

Methods: We evaluated the prognostic accuracy of the ALBI grade in estimating overall survival (OS) in a large, multi-center study including 2426 patients, including a large proportion of intermediate-stage patients treated with chemoembolization (n=1461) accrued from Europe, the United States and Asia.

Results: Analysis of survival by primary treatment modality confirmed the ALBI grade as a significant predictor of patient OS after surgical resection (p<0.001), trans-arterial chemoembolization (p<0.001) and sorafenib (<0.001). Stratification by Barcelona Clinic Liver Cancer stage confirmed the independent prognostic value of the ALBI across the diverse stages of the disease, geographical regions of origin and time of recruitment to the study (p<0.001).

Conclusions: In this large, multi-center retrospective study, the ALBI grade satisfied the criteria for accuracy and reproducibility following statistical validation in Eastern and Western HCC patients, including those treated with chemoembolization. Consideration should be given to the ALBI grade as a stratifying biomarker of liver reserve in routine clinical practice.

Introduction.

The mortality from hepatocellular carcinoma (HCC), the third most lethal malignancy on a global scale, is increasing despite best diagnostic and therapeutic efforts1[].

In contrast with most solid tumors, the prognosis of patients with HCC is not solely influenced by the extent of anatomic spread of the cancer but equally, if not more importantly, by the degree of functional impairment that accompanies liver cirrhosis2[].

A wide variety of prognostic algorithms have been proposed over the years, aimed at overcoming the limitations of Tumor-Node-Metastasis classification. These have variably integrated clinical domains shown to be harbingers of worse clinical outcome3[], including tumor mass and extent, liver dysfunction, poor performance status and blood alpha-fetoprotein (AFP) levels, however they have variable reported accuracy for survival prediction across early to advanced-stage HCC. Most models quantify liver functional status using the Child-Turcotte-Pugh (CTP) score, a classification originally devised in 1964 and later modified to estimate the risk of perioperative mortality in patients considered for surgical porto-systemic shunting.

Whilst having been widely adopted as a convenient, non-invasive indicator of liver function, the CTP score is limited by several constraints. Firstly, not all the constituents of the score have equal accessibility and reproducibility. The presence of ascites often requires ultrasound confirmation, whilst the assessment of minimal encephalopathy can be clinically challenging and subjective4[]. Secondly, the cut-off points pre-fixed for objective laboratory variables including albumin, bilirubin and pro-thrombin time are arbitrary, so that patients at the extremes of the distribution are classified equally as patients with marginally deranged laboratory parameters, producing so called “floor” and “ceiling effects”, which ultimately limit accurate prognostication5[]. Thirdly, apart from ascites grade, the CTP score does not include platelet counts or other biomarkers to indicate portal hypertension, a highly lethal complication of cirrhosis that can coexist within each CTP functional class6[]. The importance of diagnosing portal hypertension in HCC strongly emerges from the Barcelona Clinic Liver Cancer (BCLC) prognostic algorithm; the most widely adopted staging system of HCC in western countries, where invasive measurement of hepatic venous pressure gradients is reserved for patients who are considered for radical treatments7[].

A number of alternative non-invasive biomarkers of liver function have been proposed, including the Model for End-Stage Liver Disease (MELD) that takes into account serum creatinine, bilirubin and the international normalized PT ratio (INR) to derive a continuous score. Whilst useful in prioritizing candidates for liver transplantation8[], the MELD score has demonstrated prognostic limitations9[] and is not routinely used outside this setting, having failed to demonstrate an increased clinical utility over CTP in patients who are not amenable to liver transplantation5[]. Other markers have been evaluated as non-invasive indicators of fibrosis, initially to replace the need for histology in the assessment of severity of chronic liver disease10[]. However these only partially reproduce the synthetic and metabolic functions of the liver and the advent of transient elastography has re-defined their clinical role11

[ ADDIN EN.CITE ].

Recently, an alternative measure of liver function based solely on albumin and bilirubin, the ALBI grade, has been proposed in patients with HCC12

[ ADDIN EN.CITE ]. The ALBI grade is a prognostic nomogram emerging from the multivariate screen of routine clinico-pathologic variables in a large, international cohort of patients with HCC, further validated in a separate group of cirrhotics without cancer. Whilst appealing due to its potential to objectively dissect prognostically diverse subgroups within CTP classes, the ALBI grade was generated retrospectively and lacks external validation in independent studies: a necessary step before routine clinical use due to the risk of statistical over-fitting in determining true prognostic accuracy. Secondly, the ALBI score has not been characterized in intermediate-stage HCC, the most critical subset of patients in terms of prognostic assessment where survival is highly heterogeneous, ranging from 11 to 45 months13

[ ADDIN EN.CITE ].

The aim of this study was therefore to independently validate the prognostic value of the ALBI grade across all BCLC stages in a large collaborative study including patients from Europe, United States and Asia.

Patients and Methods.

The study population consisted of multiple, independently collected retrospective cohorts of consecutively recruited patients diagnosed with HCC either on imaging or by histologic criteria according to international guidelines14[] (Figure 1). Patients were recruited from tertiary referral centres with specialist multidisciplinary services for HCC management as part of an international research consortium to ensure adequate representation of all disease stages and aetiology of chronic liver disease. Patients who underwent liver transplantation as primary therapy for HCC were excluded. The patient population considered for this study was accrued as part of routine clinical care and was not selected amongst clinical trial participants.

Demographic data, complete blood count including liver function tests, AFP and the international normalized ratio (INR) value for pro-thrombin time were reconstructed from electronic medical records. Patients were staged using computerized tomography and/or magnetic resonance imaging as clinically required to derive the number of focal hepatic lesions and maximum tumor diameter detected during contrast enhancement. Computation of CTP functional class and BCLC stage followed standard pre-published methodology15[]. The ALBI grade was calculated using the following equation: linear predictor = (log10 bilirubin μmol/L × 0.66) + (albumin g/L × -0.085). The continuous linear predictor was further categorised into 3 different grades for prognostic stratification purposes: grade 1 (less than -2.60), grade 2 (between -2.60 and -1.39) and grade 3 (above -1.39) as previously described12

[ ADDIN EN.CITE ].

The primary clinical endpoint of the study was overall survival (OS), calculated from the date of initiation of treatment (surgery, first chemoembolization or initiation of sorafenib) to the date of death and/or last follow up. Due to the significant heterogeneity in the study population, survival analysis was stratified by treatment modality to include patients treated with curative resection and palliative patients amenable to locoregional therapies and systemic treatment with sorafenib.

Surgical resection cohort (n=587). To validate the score in early stage HCC, we utilised a cohort of 587 previously reported patients who underwent hepatic resection between April 2000 and 2012 as primary radical treatment for HCC at the Department of Gastroenterological Surgery, Dokkyo Medical University Hospital (Japan) as previously described16

[ ADDIN EN.CITE ].

Locoregional therapy (LRT) cohort (n=1461). Due to the documented survival heterogeneity in intermediate stage HCC and the diverse treatment protocols and re-treatment criteria for transarterial chemo-embolization (TACE) adopted in Europe, USA and Asia we opted to cross-validate the ALBI grade in different patient cohorts according to geographical region to fully ascertain its reproducibility: LRT-USA, LRT-Europe and LRT-Asia.

From a large, prospectively maintained database of 1202 patients treated at the University of Pittsburgh (USA) we derived a first dataset of 315 individuals with biopsy-proven HCC who were not candidates for surgical resection, RFA or hepatic transplantation and who received conventional selective cisplatin-TACE from 1989 to 200517[].

The European dataset of 423 patients included 297 subjects (70%) who underwent conventional frontline selective/superselective TACE at the University Hospital Freiburg (Germany) between 2003 and 2015; a second group of 64 treated at Imperial College London (UK) between 2001 and 2012 and a third subgroup of 62 patients from the academic Liver Unit in Novara (Italy), treated between 2004 and 2013.

A larger Asian dataset of 723 patients combined 79 (11%) from St. Mary’s Hospital Catholic University of Korea at Incheon (Republic of South Korea), prospectively recruited between 06/2011 and 07/2012, and a further 644 (89%) consecutive patients with unresectable HCC treated with TACE at the Kinki University Faculty of Medicine (Japan) between 01/2004 and 08/2013. Conventional selective or superselective TACE was administered following multidisciplinary discussion as reported by Pinato et al ADDIN EN.CITE [18]. In both European and Asian datasets radiologic response to TACE based followed the modified Response Evaluation Criteria in Solid Tumors (mRECIST) ADDIN EN.CITE [19] on contrast-enhanced CT scan, 6-8 weeks post-TACE. None of the patients included in the Asian dataset were part of the original ALBI qualification study.

Sorafenib cohort (n=378). A multicentre dataset of 378 patients with advanced HCC receiving sorafenib between 2008-2015 was constructed including 147 from 3 European institutions including 56 (15%) from Imperial College London (UK), 51 (13%) from the Academic Liver Unit in Novara (Italy) and 41 (11%) from the University Hospital in Freiburg (Germany). These patients were merged with a larger dataset of 230 subjects (61%) with similar characteristics recruited from the Kinki University Faculty of Medicine (Japan).

Statistical Analysis.

Univariate analysis of the different clinical factors associated with survival was performed using Kaplan-Meier curves, with differences in OS between each stratum being tested using Log-rank statistics. The independent prognostic value of each factor was further tested using by Cox proportional hazards regression models ADDIN EN.CITE [16]. Formal assessment of the proportional-hazard assumption by means of log-likelihood ratio tests over time-bands yielded no evidence that the effects of the ALBI and CTP scores varied over time (ALBI, p=0.078; CTP, p=0.1325). Analyses of survival were performed using SPSS package version 20.0 (IBM Inc., USA).

Harrell’s concordance index method (c-index) was used to rank the different staging systems according to their capacity of discriminating patients according to outcome. For this purpose, we assessed the effect of the candidate risk factors using the Cox model using R20[] and the Statistical Analysis System (SAS, Cary, NC, USA). We used the rms packages of Dr. Frank Harrell21[] to identify a subset of predictors by backward elimination as previously described16

[ ADDIN EN.CITE ]. Where we assessed the predictive ability of a Cox proportional hazards model, we compared the actual survival outcomes of usable pairs of patients with the values of their estimated prognostic indices from the Cox model to generate a c-index statistic. To correct for the overoptimism generated during model selection, the c-index stastic was internally validated using established bootstrapping techniques with 150 iterations.

Results.

Characteristics of the studied populations are presented in Table 1. Median OS was 54 months for the surgical cohort, whilst in the LRT cohorts survival ranged between 10 and 36 months, being worse in the LRT-USA cohort. Patient in the sorafenib cohort had a median OS of 9 months (Table 1).

Surgical resection cohort.

Of the 587 patients treated with liver resection, 381 (65%) satisfied BCLC-A criteria, whilst 44 patients (7%) were of BCLC-C stage due to presence of nodal spread in 7 patients (1%) and performance status of 1 in 38 (6%). The most prevalent aetiology was HCV infection (n=395, 67%) followed by HBV (n=146, 25%). Median follow up time was 41 months (6-143 months) and the overall event rate was 54%. The majority of patients were of CTP class A (n=477, 81%).

When considering OS classified according to the ALBI grade median OS was 82 months (range 58-105) in patients with ALBI grade 1 (n=127, 22%), 50 months (44-55) for ALBI 2 (n=435, 77%) and 42 months for ALBI 3 (n=25, 4%), (HR 1.7, 95%CI 1.3-2.2, p<0.001) (Figure 2A). We then considered solely patients with CTP A (n=477), who are optimal candidates for resection, and confirmed ALBI grade as a strong predictor of OS in this patient group with a median OS of 82 months (range 58-105) for ALBI 1 (n=127, 26%) and 51 months (range 45-56) for ALBI 2 (n=350, 74%) (HR 1.7, 95%CI 1.3-2.2, p<0.001) (Figure 2B).

Multivariable analysis of survival confirmed advanced ALBI as a significant predictor of worse OS in a Cox regression model adjusted for BCLC stage. Individual hazard ratios are presented in Table 2.

The predictive ability of the ALBI grade was tested in comparison with CTP by means of Harrell’s concordance index. ALBI displayed an overall better discriminatory ability in predicting OS with a c-score of 0.57 (95%CI 0.54-0.59) when considered alone and 0.67 (95%CI 0.63-0.70) following adjustment by BCLC stage, compared to CTP class (c-score 0.54, 95%CI 0.54-0.59).

Locoregional therapy cohort.

The 315 patients from the United States were treated with conventional TACE as described before17[], the majority were BCLC-B (n=272, 86%) and within CTP class A (n=220, 70%). Median follow-up was 10 months (range 1-151 months). Most patients were cirrhotic (n=238, 76%) with alcohol excess (n=221, 70%), HCV (n=113, 40%) and HBV infection (n=81, 28%) being the most prevalent risk factors. Stratification of OS by ALBI showed a deterioration in median OS from 15.4 months (range 11-20 months) in grade 1 (n=41, 13%), to 11 months (range 8.3-14 months) in grade 2 (n=209, 66%) and 4.5 months (range 3-6 months) in grade 3 (n=65, 21%) (HR 1.4, 95%CI 1.1-1.7, p=0.002) (Figure 2C). As shown in Figure 2C and in Table 2, we found no statistically significant difference in OS between patients of ALBI grade 1 and 2 in this patient cohort.

The majority of LRT-Europe cohort (n=423) received TACE within BCLC-B staging criteria (n=310, 73%) and CTP A (n=307, 72%) with the remaining 113 exceeding these criteria due to segmental PVT not contra-indicating TACE (n=40, 9%), extra-hepatic spread (n=1, 0.2%) or PS=1 (n=71, 17%). Alcohol excess was the most prevalent etiologic factor (n=160, 38%) followed by HCV infection (n=115, 27%). Median follow up time was 27 months (range 1-162 months), with a mortality rate of 72%. Patient stratification according to ALBI revealed grade 1 (n=139, 33%) predicted for median survival of 39 months (range 33-44 months) compared to 18 months (range 14.3-21.6) for ALBI 2 (n=251, 60%) and 18 months for ALBI 3 (n=30, 7%) (range 14.8-21.1, HR 1.4, 95%CI 1.2-1.8, p<0.001) with no significant difference between grades 2 and 3 (Figure 2D).

Staging characteristics were similar in the LRT-Asia cohort of 723 patients, 660 (91%) being within BCLC-B stage and 67% CTP A (n=488) with the exception 63 subjects who exceeded intermediate-stage due to visceral metastatic spread (n=29, 4%) and/or segmental portal vein involvement (n=59, 8%). HCC was mostly HCV (n=454, 70%) or HBV-related (127, 18%). Patients with ALBI grade 1 (n=156, 22%) had a median OS of 51.8 months (range 45-57 months) compared to 34 months (range 30-39 months) of ALBI 2 (n=482, 67%) and 21 (range 14-28 months) of ALBI 3 (n=85, 12%)(HR 1.7, 1.4-2.0, p<0.001) (Figure 2E).

To verify its independent prognostic value, the ALBI grade was jointly tested with other prognostic factors including CTP and BCLC stage in alternative multivariate Cox regression model. The effect of ALBI in predicting patients’ OS across each of the LRT cohorts is reported in Table 2.

The discriminatory capacity of the ALBI grade as calculated by the corresponding c-index score was substantially homogeneous across each LRT cohort, ranging from 0.57 to 0.58 when ALBI was considered alone, and 0.58-0.65 when adjusted for BCLC stage. Individual c-index scores for ALBI and CTP are presented in Table 2.

Sorafenib cohort.

Sorafenib was prescribed to 229 patients (61%) in BCLC-C stage, 280 (74%) within CTP class A. Etiology of chronic liver disease included HCV (n=161, 43%), HBV infection (n=62, 16%) and alcohol excess (n=56, 15%). Patients were followed up for a median of 13 months (range 1-74 months). There were 264 deaths (70%) in total. Median duration of treatment was 2.8 months (range 1-70 months). At time of analysis, 16 patients were actively receiving sorafenib (4%), whilst the remaining had discontinued due to radiologic progression (n=181, 48%), unacceptable toxicity (n=105, 28%), death (n=28, 4%) or other causes (n=48, 13%). The most common adverse events, occurring at any grade, were diarrhea (n=134, 35%) and palmar-plantar erythrodysesthaesia (n=131, 35%) requiring permanent dose modifications in 87 patients (23%). Periodic restaging during sorafenib therapy according to mRECIST criteria was available in 222 patients. Best radiologic response included disease progression (n=119, 31%), stable disease (n=65, 17%), partial (n=31, 8%) and complete response (n=7, 1%). When categorized according to ALBI patients with grade 1 had a median OS of 16 months (range 12-20) compared to 7.6 of grade 2 (range 5.9-9.3) and 4.8 of grade 3 (range 3-6.6) (HR 1.6, 95%CI 1.3-2.0, p<0.001) (Figure 2F).

Multivariable analysis of survival confirmed ALBI as an independent predictor of patients’ OS in advanced disease (Table 2).

The prognostic relationship between ALBI and BCLC: interaction with time.

We further confirmed the prognostic validity of the ALBI grade in a pooled analysis of patients belonging to the entire study population (n=2426). As shown in Table 3, the ALBI grade was confirmed as a predictor of OS following adjustment for BCLC stage. Given the wide accrual times (1989-2015) we performed further analyses of survival by including an interaction term between the scores of interest (either ALBI or CTP) and start time of therapy, dichotomized as prior to or after year 2000, to reflect subsequent adoption of the BCLC treatment allocation algorithm. Whilst there was no evidence of a change in the effect of the CTP score on OS before and after year 2000 (p=0.5982), there was some evidence of a variation in the effect of the ALBI score alone (p=0.0031) and in combination with BCLC (p=0.0637). The estimates of these two models stratified by time are presented in Table 4. Stratified analyses provide evidence of a predictive effect of the ALBI score on survival both in the pre- and post year 2000 datasets, both in unadjusted and adjusted analyses, although the effects of the score were greater in the post year 2000 dataset.

Discussion

HCC arises as a complication of chronic liver disease and cirrhosis in the vast majority of cases. The prognostic interaction between progressive malignancy and underlying liver functional impairment poses a unique challenge in the provision of optimal treatment strategies in the individual patient ADDIN EN.CITE [22]. The geographical distribution of different etiologic factors and the diverse accessibility to various treatment modalities are amongst the factors that make the clinical course of HCC particularly variable across countries and institutions23[]. This is particularly true for intermediate-stage HCC, where the variation in tumor burden, hepatic reserve combined with the recognized heterogeneity in loco-regional treatment schedules exert a crucial impact on patients’ survival, which recognizes a >20 months variation in this patient subpopulation24[]. The acknowledged clinical heterogeneity of HCC suggests the need for objective, reproducible and accurate prognostic biomarkers to improve patient outcomes25[].

In our large, collaborative, multi-institutional study we sought to independently validate the prognostic ability of the ALBI grade, a recently qualified nomogram suggested to more accurately predict patients’ mortality by solely combining albumin and bilirubin levels without the need for subjective determinants of liver failure including ascites and encephalopathy.

Using a stage-stratified approach, we investigated the prognostic ability of the ALBI in both curative and palliative settings, demonstrating adequate and clinically meaningful stratifying potential for the score across all the BCLC stages of the disease.

In early-stage HCC, we demonstrated that the ALBI grade could predict worse prognosis within CTP A class patients, who are universally recognized as optimal candidates for surgical resection26[]. Whilst our findings in early-stage disease are retrospective and based on a single-center experience, they corroborate previous evidence showing that the ALBI may overcome the prognostic limitations of the CTP classification, allowing for a more subtle quantification of hepatic reserve in patients with grossly preserved synthetic function, perhaps by abolishing “floor” and “ceiling” effects applying to CTP class computation. Clinically, in fact, these patients rarely display evidence of significant coagulopathy, refractory ascites or encephalopathy, a point that strengthens the relevance of pre-operative albumin and bilirubin in the prognostic assessment of surgical candidates27[].

An increasing number of studies have tested the prognostic accuracy of the ALBI grade in multi-institutional cohorts and recently proposed its integration with routinely used prognostic models including TNM28[] and BCLC ADDIN EN.CITE [29]. However, in a recent study by Hiraoka, which only included Japanese patients, representation of BCLC-B patients accounted for only 12% of the total of patients studied28[], and similar figures apply to the Chan study where a high proportion of patients (up to 33%) received best supportive care only, a factor that has influenced survival outcomes ADDIN EN.CITE [29].

To our knowledge, this is the first study to provide evidence of the clinical utility of the ALBI grade in a large, multi-institutional cohort of patients who received TACE for unresectable HCC, largely within BCLC-B criteria. We confirmed the ALBI as an accurate predictor of patient’s survival in this patient group. Due to the expected heterogeneity in patients’ survival stemming from both patient features and treatment schedules across continents, we opted to validate the prognostic ability of the ALBI separately by constructing 3 geographically distinct patient cohorts characterized by diverse ethnicity and underlying etiology. Unsurprisingly, despite relatively balanced tumor staging features and CTP class, survival was significantly different across the 3 studied cohorts, being worse in US patients compared to the European and Asian cohorts.

Whilst median survival for the European and Asian cohorts is in line with the life expectancy of patients currently presenting with BCLC-B HCC24[], American patients were mostly treated prior to the dissemination of the BCLC algorithm in the early 2000s: although reconstructed retrospectively by medical records review, the algorithm did not guide treatment allocation in these patients, with inherent implications in terms of survival15[]. It is also likely that an equally important influence on survival might have derived from the evolving technical improvements in the delivery of TACE within the accrual period (1989-2005). Stratified analyses according to time provide evidence that the prognostic role of the ALBI grade was preserved in patients assessed prior to and following the dissemination of the BCLC treatment guidelines. Interestingly, though, the effect of the score in influencing patients’ survival was greater in the post year 2000 dataset, a finding that might correlate with an improved patient selection and more homogeneous management over time (Table 4).

Whilst these considerations over patients’ homogeneity should be acknowledged when interpreting our results, the preserved accuracy of the ALBI in each of the studied cohorts further strengthens its independence as a prognostic predictor in HCC despite heterogeneity in ethnicity, accrual period, disease etiology, patient selection and management strategies. Crucially, in our study, the ALBI grade displayed comparable accuracy to CTP in estimating mortality prior to TACE suggesting that combination of pre-treatment albumin and bilirubin are indeed the strongest predictors of the CTP classification. This is not a novel finding in intermediate-stage HCC, where a number of scores relying on pre-treatment albumin and bilirubin levels have been proposed to guide TACE administration in HCC, with the intent of sparing patients with poor prognostic features from potentially futile loco-regional therapies and migrate them to systemic treatment or best supportive care ADDIN EN.CITE [30-32]. Despite having potential clinical utility, our experience suggests that patient heterogeneity and differences in cut-off values for variables including bilirubin ADDIN EN.CITE [18], may influence the prognostic performance of these algorithms, a limitation that is not shared by the ALBI grade where albumin and bilirubin are considered as continuous variables, preserving their prognostic accuracy in full.

Our study also demonstrates that the ALBI grade is a prognostic maker in advanced HCC, a patient population with limited life expectancy and restricted treatment options. Based on our results, the ALBI grade was superior to CTP in identifying patients at risk of early mortality prior to sorafenib initiation as well as an independent predictor of OS. These features are of greater consequence in advanced HCC, a patient population at the focus of intense research efforts. The post-registrative extension of sorafenib therapy to early stage CTP B patients has in fact demonstrated that patients with intermediate derangement in their liver biochemistry can still benefit from treatment ADDIN EN.CITE [33] in absence of excessive toxicities34[]. In this context, a more accurate index of liver function such as the ALBI grade might optimize safe administration of experimental therapies in advanced HCC, therefore facilitating efficacy testing of investigational medicinal compounds.

In conclusion, this study provides compelling evidence that the ALBI grade is a validated prognostic index across all BCLC stages of disease. Because it relies on two single inexpensive and widely available laboratory parameters, the ALBI grade is more objective and can be rapidly computed at the bedside, without the need for special tests.

With c-index values showing moderate discriminative ability in predicting mortality from deteriorating liver function across each stage of HCC, the ALBI grade emerges as an equal and, in some instances, superior biomarker to routinely available liver functional staging systems such as the CTP class. The prognostic performance of the ALBI grade is particularly appealing in intermediate-stage disease, where we have shown it to be a widely generalizable biomarker able to overcome the heterogeneity of survival of BCLC-B stage.

The retrospective nature of our study and the relatively wide accrual times characterizing some of the recruited cohorts should be acknowledged as limitations. However, our survival analysis is robustly built on a process of independent cross-validation in large multi-institutional cohorts, which limits the potential for selection bias.

In summary, this study has validated the ALBI grade as an objective, inexpensive, readily available stratifying biomarker of poor liver reserve in HCC. Consideration should be given to its prospective validation in future clinical studies to facilitate its use in routine clinical practice.

References.

[1]Jemal A, Siegel R, Xu J, Ward E. Cancer statistics, 2010. CA: a cancer journal for clinicians 2010;60:277-300.

[2]Schutte K, Bornschein J, Malfertheiner P. Hepatocellular carcinoma--epidemiological trends and risk factors. Dig Dis 2009;27:80-92.

[3]Llovet JM, Bruix J. Novel advancements in the management of hepatocellular carcinoma in 2008. J Hepatol 2008;48 Suppl 1:S20-37.

[4]Durand F, Valla D. Assessment of prognosis of cirrhosis. Semin Liver Dis 2008;28:110-122.

[5]Cholongitas E, Papatheodoridis GV, Vangeli M, Terreni N, Patch D, Burroughs AK. Systematic review: The model for end-stage liver disease--should it replace Child-Pugh's classification for assessing prognosis in cirrhosis? Alimentary pharmacology & therapeutics 2005;22:1079-1089.

[6]Procopet B, Tantau M, Bureau C. Are there any alternative methods to hepatic venous pressure gradient in portal hypertension assessment? J Gastrointestin Liver Dis 2013;22:73-78.

[7]Cucchetti A, Cescon M, Golfieri R, Piscaglia F, Renzulli M, Neri F, et al. Hepatic venous pressure gradient in the preoperative assessment of patients with resectable hepatocellular carcinoma. J Hepatol 2015.

[8]Wiesner R, Edwards E, Freeman R, Harper A, Kim R, Kamath P, et al. Model for end-stage liver disease (MELD) and allocation of donor livers. Gastroenterology 2003;124:91-96.

[9]Cholongitas E, Marelli L, Shusang V, Senzolo M, Rolles K, Patch D, et al. A systematic review of the performance of the model for end-stage liver disease (MELD) in the setting of liver transplantation. Liver Transpl 2006;12:1049-1061.

[10]Lurie Y, Webb M, Cytter-Kuint R, Shteingart S, Lederkremer GZ. Non-invasive diagnosis of liver fibrosis and cirrhosis. World J Gastroenterol 2015;21:11567-11583.

[11]Singh S, Fujii LL, Murad MH, Wang Z, Asrani SK, Ehman RL, et al. Liver stiffness is associated with risk of decompensation, liver cancer, and death in patients with chronic liver diseases: a systematic review and meta-analysis. Clin Gastroenterol Hepatol 2013;11:1573-1584 e1571-1572; quiz e1588-1579.

[12]Johnson PJ, Berhane S, Kagebayashi C, Satomura S, Teng M, Reeves HL, et al. Assessment of liver function in patients with hepatocellular carcinoma: a new evidence-based approach-the ALBI grade. J Clin Oncol 2015;33:550-558.

[13]Dufour JF, Bargellini I, De Maria N, De Simone P, Goulis I, Marinho RT. Intermediate hepatocellular carcinoma: current treatments and future perspectives. Ann Oncol 2013;24 Suppl 2:ii24-29.

[14]European Association For The Study Of The L, European Organisation For R, Treatment Of C. EASL-EORTC clinical practice guidelines: management of hepatocellular carcinoma. J Hepatol 2012;56:908-943.

[15]Llovet JM, Bru C, Bruix J. Prognosis of hepatocellular carcinoma: the BCLC staging classification. Semin Liver Dis 1999;19:329-338.

[16]Pinato DJ, Stebbing J, Ishizuka M, Khan SA, Wasan HS, North BV, et al. A novel and validated prognostic index in hepatocellular carcinoma: the inflammation based index (IBI). J Hepatol 2012;57:1013-1020.

[17]Carr BI, Kondragunta V, Buch SC, Branch RA. Therapeutic equivalence in survival for hepatic arterial chemoembolization and yttrium 90 microsphere treatments in unresectable hepatocellular carcinoma: a two-cohort study. Cancer 2010;116:1305-1314.

[18]Pinato DJ, Arizumi T, Allara E, Jang JW, Smirne C, Kim YW, et al. Validation of the hepatoma arterial embolization prognostic score in European and Asian populations and proposed modification. Clin Gastroenterol Hepatol 2015;13:1204-1208 e1202.

[19]Lencioni R, Llovet JM. Modified RECIST (mRECIST) assessment for hepatocellular carcinoma. Semin Liver Dis 2010;30:52-60.

[20]Team RC. R: A language and environment for statistical computing. Vienna: R Foundation for Statistical Computing; 2014.

[21]Harrell FE. Regression Modeling Strategies. 3.0 ed; 2010.

[22]Liu PH, Hsu CY, Hsia CY, Lee YH, Su CW, Huang YH, et al. Prognosis of Hepatocellular Carcinoma: Assessment of Eleven Staging Systems. J Hepatol 2015.

[23]de Lope CR, Tremosini S, Forner A, Reig M, Bruix J. Management of HCC. J Hepatol 2012;56 Suppl 1:S75-87.

[24]Sangro B, Salem R. Transarterial chemoembolization and radioembolization. Semin Liver Dis 2014;34:435-443.

[25]Bolondi L, Burroughs A, Dufour JF, Galle PR, Mazzaferro V, Piscaglia F, et al. Heterogeneity of patients with intermediate (BCLC B) Hepatocellular Carcinoma: proposal for a subclassification to facilitate treatment decisions. Semin Liver Dis 2012;32:348-359.

[26]Llovet JM, Schwartz M, Mazzaferro V. Resection and liver transplantation for hepatocellular carcinoma. Semin Liver Dis 2005;25:181-200.

[27]Sacco R, Antonucci M, Bresci G, Corti A, Giacomelli L, Mismas V, et al. Curative therapies for hepatocellular carcinoma: an update and perspectives. Expert Rev Anticancer Ther 2015.

[28]Hiraoka A, Kumada T, Michitaka K, Toyoda H, Tada T, Ueki H, et al. Usefulness of albumin-bilirubin (ALBI) grade for evaluation of prognosis of 2584 Japanese patients with hepatocellular carcinoma. J Gastroenterol Hepatol 2015.

[29]Chan AW, Kumada T, Toyoda H, Tada T, Chong CC, Mo FK, et al. Integration of albumin-bilirubin (ALBI) score into Barcelona clinic liver cancer (BCLC) system for hepatocellular carcinoma. J Gastroenterol Hepatol 2016.

[30]Kadalayil L, Benini R, Pallan L, O'Beirne J, Marelli L, Yu D, et al. A simple prognostic scoring system for patients receiving transarterial embolisation for hepatocellular cancer. Ann Oncol 2013;24:2565-2570.

[31]Pinato DJ, Karamanakos G, Arizumi T, Adjogatse D, Kim YW, Stebbing J, et al. Dynamic changes of the inflammation-based index predict mortality following chemoembolisation for hepatocellular carcinoma: a prospective study. Alimentary pharmacology & therapeutics 2014;40:1270-1281.

[32]Hucke F, Pinter M, Graziadei I, Bota S, Vogel W, Muller C, et al. How to STATE suitability and START transarterial chemoembolization in patients with intermediate stage hepatocellular carcinoma. J Hepatol 2014;61:1287-1296.

[33]Pressiani T, Boni C, Rimassa L, Labianca R, Fagiuoli S, Salvagni S, et al. Sorafenib in patients with Child-Pugh class A and B advanced hepatocellular carcinoma: a prospective feasibility analysis. Ann Oncol 2013;24:406-411.

[34]LG DAF, Barroso-Sousa R, Bento AD, Blanco BP, Valente GL, Pfiffer TE, et al. Safety and efficacy of sorafenib in patients with Child-Pugh B advanced hepatocellular carcinoma. Mol Clin Oncol 2015;3:793-796.

Figure Legends.

Figure 1. Study flow diagram illustrating patient disposition across the studied cohorts.

Figure 2. Kaplan Meier curve analysis showing the effect of the ALBI grade as predictor of overall survival across each primary treatment modality in HCC: surgically resected patients (Panel A), surgically resected patients within Child-Turcotte-Pugh Class A criteria (Panel B), patients treated with locoregional therapy from USA (Panel C), Europe (Panel D) and Asia (Panel E) and patients treated with sorafenib (Panel F).

Acknowledgements:

DJP is supported by the National Institute for Health Research (NIHR) as well as grant funding from Action Against Cancer, the Academy of Medical Sciences and the Imperial NIHR Biomedical Research Centre (BRC). The authors would like to acknowledge Dr. Les Huson for his statistical support. This work was presented orally by DJP at the EASL 2016 International Liver Meeting, Barcelona.

1