Alimentary Pharmacology and Therapeutics

The albumin–bilirubin grade improves hepatic reserve

estimation post-sorafenib failure: implications for drug
development
D. J. Pinato*, C. Yen*, D. Bettinger†, R. Ramaswami*, T. Arizumi‡, C. Ward*, M. Pirisi§,¶, M. E. Burlone§, R. Thimme†,
M. Kudo‡ & R. Sharma*

*Department of Surgery & Cancer, SUMMARY

Hammersmith Campus of Imperial
College London, London, UK.
†
Department of Medicine II,
Background
University Hospital Freiburg, Freiburg, Drug development in hepatocellular carcinoma (HCC) is limited by disease
Germany. heterogeneity, with hepatic reserve being a major source of variation in sur-
‡
Department of Gastroenterology and vival outcomes. The albumin–bilirubin (ALBI) grade is a validated index of
Hepatology, Kindai University School
liver function in patients with HCC.
of Medicine, Osaka, Japan.
§
Department of Translational
Medicine, Universita degli Studi del Aim
Piemonte Orientale “A. Avogadro”, To test the accuracy of the ALBI grade in predicting post-sorafenib overall
Novara, Italy.
¶
survival (PSOS) in patients who permanently discontinued treatment.
Interdisciplinary Research Center of
Autoimmune Diseases, Universita
degli Studi del Piemonte Orientale “A. Methods
Avogadro”, Novara, Italy. From a prospectively maintained international database of 447 consecutive
referrals, we derived 386 eligible patients treated with sorafenib within Barce-
lona Clinic Liver Cancer C stage (62%), 75% of whom were of Child class A
Correspondence to:
Dr R. Sharma, Department of Surgery
at initiation. Clinical variables at sorafenib discontinuation were analysed for
& Cancer, Imperial College London their impact on post-sorafenib overall survival using uni- and multivariable
Hammersmith Campus, Du Cane analyses.
Road, W12 0HS London, UK.
E-mail: r.sharma@imperial.ac.uk Results
Median post-sorafenib overall survival of the 386 eligible patients was
Publication data 3.4 months and median sorafenib duration was 2.9 months, with common-
Submitted 1 October 2016 est causes of cessation being disease progression (68%) and toxicity (24%).
First decision 20 October 2016 At discontinuation, 92 patients (24%) progressed to terminal stage, due to
Resubmitted 20 October 2016
worsening Child class to C in 40 (10%). Median post-sorafenib overall sur-
Resubmitted 23 November 2016
Accepted 24 November 2016
vival in patients eligible for second-line therapies (n = 294) was 17.5, 7.5
EV Pub Online 24 January 2017 and 1.9 months according respectively to ALBI grade 1, 2 and 3
(P < 0.001).
The Handling Editor for this article was
Professor Peter Hayes, and it was
Conclusions
accepted for publication after full peer-
review. The ALBI grade at sorafenib discontinuation identifies a subset of patients
with prolonged stability of hepatic reserve and superior survival. This may
allow improved patient selection for second-line therapies in advanced HCC.

Aliment Pharmacol Ther 2017; 45: 714–722

714 ª 2017 John Wiley & Sons Ltd

doi:10.1111/apt.13904

INTRODUCTION
In patients with hepatocellular carcinoma (HCC), liver
cirrhosis carries an independent clinical course from that
of neoplastic disease progression1 and often represents a
barrier to the provision of active anti-cancer treatment.2
This is particularly true in the more advanced stages of
the disease, where liver dysfunction together with deteri-
orating performance status (PS)3 remain the strongest
adverse prognostic factors.4, 5
Sorafenib remains the only first-line treatment option
of proven survival benefit in advanced HCC.6, 7 How-
ever, depth and length of response to sorafenib are mod-
est, with <2% objective response rates. Even in patients
who achieve initial disease control, progression inex-
orably occurs within a median interval of 3 months.8
Pre-registrative clinical trials of sorafenib in HCC
were originally conducted in highly selected populations,
as evidenced by high screening failure rates approaching
30%.6 Whilst clinically developed in patients with opti-
mal liver function, subsequent implementation of sorafe-
nib in routine practice to include subjects with a wider
variation in hepatic reserve and PS has highlighted
numerous challenges in the clinical management of these
patients.9 In the clinic, neoplastic disease progression,
deteriorating PS, hepatic decompensation or unaccept-
able treatment-related toxicity are the leading causes of
permanent sorafenib discontinuation.10, 11 At treatment
cessation, the majority of patients will only qualify for
best supportive care, whereas a smaller proportion might
be suitable for second-line therapies, an area that is cur-
rently at the focus of intense clinical research efforts.12
Adequate hepatic function is a universal pre-requisite
to access experimental therapies. However, it has been
recently suggested that the prognostic ability of the Child
class, the scoring system traditionally employed to evalu-
ate liver function based on five parameters, might be
outclassed by a novel, more objective biochemical nomo-
gram solely composed of serum albumin and bilirubin,
the ALBI grade.13 The improved clinical utility of this
score stems from its ability to further dissect patient
groups with different survival outcomes within each
Child functional class14 as well as Barcelona Clinic Liver
Cancer (BCLC) stage of HCC.15
More accurate prognostic predictors of liver functional
reserve are highly required in patients who fail sorafenib
treatment, given that a proportion of patients might be
eligible for further lines of anti-cancer treatment in an
experimental setting. Whilst previous studies have
focused on the radiological pattern of progression on
Aliment Pharmacol Ther 2017; 45: 714–722
ª 2017 John Wiley & Sons Ltd
Albumin-bilirubin grade predicts post-sorafenib survival
sorafenib as a predictor of survival, this trait, while bio-
logically important and potentially useful as a stratifica-
tion factor in second-line studies, does not influence
patients’ eligibility to further lines of therapy, for which
the main barriers are represented by progression to ter-
minal-stage disease either due to worsening PS or liver
decompensation.16
Furthermore, as an improved marker of functional
reserve, the ALBI grade might be useful to facilitate
patient stratification according to their risk of toxicity
from systemic anti-cancer treatment, a non-negligible
concern in patients with underlying impairment of hep-
atic drug clearance.17
We therefore conducted a study to ascertain the prog-
nostic value of the ALBI grade in patients who perma-
nently discontinued sorafenib treatment in a large
collaborative study including patients from Europe, and
Asia.
PATIENTS AND METHODS
The study population derives from a prospectively main-
tained, multi-centre patient dataset including 447 consec-
utive patients treated with sorafenib between 2008 and
2015 within an international research consortium. We
excluded 30 patients who were on sorafenib and further
39 due to incomplete data (Figure 1). The final dataset
(n = 386) comprised 191 patients from three European
institutions including 76 (20%) treated at Imperial Col-
lege London (UK), 44 (11%) from the Academic Liver
Unit in Novara (Italy) and 71 patients (18%) treated at
the University Hospital in Freiburg (Germany). These
patients were incorporated into a larger dataset of 195
subjects (51%) with similar features treated at the Kindai
University Faculty of Medicine in Osaka (Japan).
The primary aim was to evaluate the ALBI grade as a
predictor of post-sorafenib overall survival (PSOS) and
verify its accuracy in comparison with Child class. PSOS
was calculated from the time of sorafenib discontinua-
tion to the time of death/last-documented follow-up.
Post-sorafenib treatment status was documented for all
patients.
The ALBI was calculated using the formula: linear pre-
dictor = (log10 bilirubin 9 0.66) + (albumin 9 0.085),
where bilirubin is expressed in lmol/L and albumin in g/
L. Patients were categorised as ALBI grade 1 if the linear
predictor ≤ 2.60; ALBI grade 2 if more than 2.60 to
≤ 1.39 and ALBI grade 3 if > 1.39.13
All patients were diagnosed with HCC in accordance
to international guidelines.18 Sorafenib was administered
715
D. J. Pinato et al.
Imperial College
(London, UK)
98
University Hospital Piemonte Orientale University Kindai University
(Freiburg, Germany) (Novara, Italy) (Osaka, Japan)
71 51
69 patients excluded:
• Missing data (39)
• Ongoing sorafenib therapy (30)
Study Population
386
in accordance to the BCLC treatment algorithm to stage
C patients or earlier if, on the basis of a multi-disciplin-
ary discussion, they were deemed unfit for or had failed
prior loco-regional therapies. Routine clinical reviews
during sorafenib therapy were performed at each cycle of
treatment and included physical examination, blood test
review and adverse events evaluation. Reporting of sora-
fenib-related AEs followed NCI Common Terminology
Criteria for Adverse Events (CTCAE) version 3.19
Clinical data at sorafenib commencement and discon-
tinuation included complete blood count, liver function
tests, alpha-fetoprotein (AFP) levels and the international
normalised ratio value for prothrombin time. Radiologi-
cal staging of the disease was performed using comput-
erised tomography (CT) and/or magnetic resonance
imaging as clinically required. Computation of the ALBI
grade,13 Child functional class and BCLC stage followed
standard pre-published methodology.20 Periodic restaging
using triple-phase CT during sorafenib treatment was
performed at 8–12 weekly intervals.21 Reasons for per-
manent sorafenib cessation included disease progression,
unacceptable toxicity defined by grade 2–4 AEs not
responding to dose reductions and/or temporary sorafe-
nib discontinuation. In patients with multiple causes for
treatment discontinuation (i.e. liver decompensation or
unacceptable toxicity), tumour progression was consid-
ered the primary cause if confirmed radiologically.10 The
study was conducted in accordance to Good Clinical
Practice standards following the ethical guidelines pub-
lished in the Declaration of Helsinki.
Statistical analysis
Univariable analysis survival followed Kaplan–Meier
methodology and log-rank statistics. Cox proportional
hazards regression models were for multivariable
716
227
Figure 1 | Study flow chart.
analyses of survival, following formal assessment of the
proportional-hazard assumption for each variable using
log-likelihood ratio tests over time.22 Receiver operating
characteristic (ROC) methodology was used to test the
accuracy of the candidate biomarkers in predicting 90-
days mortality.23 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. Where we assessed the predictive
ability of a Cox proportional hazards model, we com-
pared the actual survival outcomes of usable pairs of
patients with the values of their estimated prognostic
indices from the Cox model. Where the assessment of
prediction of multiple biomarkers was performed, the c-
index was adjusted within the R package for the over-
optimism produced by modelling, and assessment being
done on the same data via comparison with 150 boot-
strap samples. For all analyses, a P ≤ 0.05 was consid-
ered statistically significant. Statistical analyses were
performed using SPSS package version 20.0 (IBM Inc.,
Chicago, IL, USA) and R Statistical Computing Environ-
ment (R Foundation, Vienna, Austria).
RESULTS
Patient characteristics
Patient features are presented in Table 1. The majority
of patients were of BCLC-C stage (n = 248, 62%). The
predominant aetiological factors of chronic liver disease
were HCV infection (n = 150, 39%), HBV infection
(n = 62, 16%) and alcohol excess (n = 72, 19%). In total,
290 patients (75%) Child class A at sorafenib initiation,
whereas 96 (25%) were of Child B class. All patients had
a PS ≤ 1 at sorafenib initiation. Sorafenib discontinuation
followed primarily progression of disease in 263 patients
Aliment Pharmacol Ther 2017; 45: 714–722
ª 2017 John Wiley & Sons Ltd
 Albumin-bilirubin grade predicts post-sorafenib survival

Table 1 | Clinical characteristics of the studied patients Table 1 | (Continued)

Characteristic N = 386 (%) Characteristic N = 386 (%)

Age in years, median (range) 71 (33–92) 0 77 (20)

Gender 1 156 (40)
Male 309 (80) 2 93 (24)
Female 77 (20) ≥3 60 (16)
Aetiology of liver disease Subsequent treatment for HCC
Viral 209 (54) Best supportive care 300 (78)
Non viral 177 (46) Further anti-cancer treatment 86 (22)
AFP (ng/mL) Sorafenib discontinuation
<400 244 (63) Progressive disease 263 (64)
>400 142 (37) Toxicity 91 (24)
Albumin (g/L), median (range) 30 (16–48) Liver failure 24 (6)
Bilirubin (lmol/L), median (range) 21 (3–439) Others 8 (2)
AST (IU/L), median (range) 75 (11–2462)
ALT (IU/L), median (range) 48 (10–1350) All clinicopathological variables refer to the moment of
INR, median (range) 1.1 (0.6–3.5) sorafenib discontinuation unless otherwise stated.
Platelet count, median (range) 141 (14–673)
Child–Turcotte Pugh Score
A5 77 (20) (64%), followed by unacceptable toxicity in 91 (24%).
A6 78 (21) The most common sorafenib-related AEs occurring at
B7 77 (21)
any grade during treatment were diarrhoea (n = 143,
B8 67 (18)
B9 38 (10) 37%), palmo-plantar erythrodysesthaesia (PPE) (n = 139,
C10 31 (8) 36%) and fatigue (n = 88, 22%). Similarly, when consid-
C11 4 (1) ering AEs ≥Grade 2, PPE was the most prevalent
C12 5 (1) (n = 33, 9%) followed by diarrhoea (n = 24, 6%) and
ALBI grade
Grade 1 27 (7)
fatigue (n = 21, 5%). In total 196 patients (51%) had
Grade 2 171 (45) experienced at least 1 Grade 2 AE during sorafenib treat-
Grade 3 188 (48) ment, whereas 104 (27%) had experienced at least 1
Tumour morphology Grade 3 AE. Toxicity prompted permanent sorafenib
Unifocal 36 (10)
dose reductions in 59 patients (15%).
Multifocal (<50% of liver replacement) 221 (57)
Massive (>50% of liver replacement) 129 (32) At the time of sorafenib discontinuation, the stage-dis-
Maximum diameter of largest lesion 4.2 (1–20) tribution according to BCLC had significantly changed
(cm), median (range) compared to baseline, with the majority of patients clus-
Extrahepatic spread tering within BCLC-C criteria (n = 189, 49%) followed
Absent 239 (63)
Present 147 (37) by BCLC-B (n = 105, 27%) and D (n = 92, 24%). Stage
Portal vein involvement migration to BCLC-D derived from progression to Child
Absent 287 (74) class to C in 39 patients (10%) and worsening PS to >2
Present 99 (26) in 53 patients (14%). Following sorafenib discontinua-
BCLC stage at sorafenib initiation
A–B 154 (38)
tion, the majority of patients received best supportive
C 248 (62) care (n = 294, 76%). Of the 92 patients (24%) who
BCLC stage at sorafenib cessation received further active treatment, 24 (6%) received >1
B 105 (27) line of therapy.
C 189 (49)
D 92 (24)
Prior treatment for HCC Post-sorafenib overall survival
Orthotopic liver transplantation 2 (1) At the time of analysis (April 2016), there were 298
Surgical resection 62 (16) deaths (77%). Patients were followed up for a median of
Radiofrequency ablation 114 (30) 13 months (range 1–70 months). Median duration of
Transarterial chemoembolisation 231 (60)
Other 53 (14)
treatment was 2.9 months (range 1–53 months). Median
Number of prior treatments lines OS from the time of sorafenib initiation was 9.4 months
(95% CI 7.6–11.0), whereas median PSOS was

Aliment Pharmacol Ther 2017; 45: 714–722 717

ª 2017 John Wiley & Sons Ltd
D. J. Pinato et al.

3.4 months (95% CI 2.5–4.3). The median PSOS consid-

ering BCLC stage at sorafenib discontinuation was
9 months (95% CI 5.7–12.3) for stage B, 3.6 months for
stage C (95% CI 2.4–4.7) and 1.1 months for stage D
(95% CI 0.5–1.6). The hazard ratio (HR) between C and
B was 2.0 (95% CI 1.4–2.7) and between D and B was
3.5 (95% CI 2.5–5.1) (log-rank P < 0.001). Patients with-
drawing sorafenib due to AEs had an improved median
survival of 5.8 months (95% CI 3.4–8.1) compared to 2.6
(95% CI 1.9–3.4) in patients who progressed on treat-
ment (HR 0.65 95% CI 0.52–0.84, log-rank P < 0.001).
We verified the relationship between a number of clini-
copathological traits at sorafenib discontinuation and
PSOS including the presence of extrahepatic spread [me-
dian PSOS 4.9 vs. 3.4 months, HR 1.4 (95% CI 1.1–1.9),
log-rank P = 0.01) and AFP >400 (median PSOS 5.0 vs.
2.9 months, HR 1.3 [95% CI 1.1–1.7], log-rank P = 0.02.

Prognostic comparison of Child class and ALBI grade

at sorafenib discontinuation
For this study, we focused on a comparative analysis of
biomarkers of liver functional reserve. The ALBI grade
was calculated in all patients. When categorised according
to ALBI, patients with grade 1 (n = 27) had a median OS
of 13 months (95% CI 1.0–25) compared to 6.3 of grade 2
(n = 171, range 4.3–8.3) and 1.6 of grade 3 (n = 188, 1.0–
2.0; log-rank P < 0.001). The HR between ALBI 2 and 1
was 1.7 (95% CI 1.1–2.9) and between ALBI 3 and ALBI 1
was 4.1 (95% CI 2.5–6.8) (Figure 2a).
Child classification was derived in 377 patients (98%).
Child class was associated with post-sorafenib survival
with median OS of 8.2 months (95% CI 5.2–11) for Child
A (n = 160), 2.1 months (95% CI 1.6–2.6) for Child B
(n = 178) and 1 month for Child C patients (n = 39, 95%
CI 0.5–1.3, log-rank P < 0.001; Figure 2b). The HR
between Child class B and A was 2.4 (95% CI 1.8–3.1) and
between Child C and A was 4.7 (95% CI 3.0–7.2). A com-
parison of ALBI grade and Child class for the 377 patients Figure 2 | Kaplan–Meier curves describing the overall
survival of patients with HCC (n = 386) at sorafenib
with both scores available is reported in Table S1.
discontinuation categorised according to albumin–
We further investigated the prognostic relationship bilirubin (ALBI) grade (a) and Child class (b).
between ALBI and Child class by evaluating the stratify-
ing potential of the ALBI across each Child class. The
analysis was limited to Child A and B classes because We performed further subgroup analyses by excluding
the entire patients with Child C were co-classified as patients who, according to current guidelines, displayed
ALBI Grade 3. As shown in Table S2, the ALBI could untreatable disease progression at sorafenib discontinua-
sub-classify patients within Child A according to signifi- tion (i.e. patients migrating to BCLC-D stage, n = 92,
cantly different survival times ranging from a median OS 24%). In a subset of 294 potential candidates for second-
of 17.5–1.0 months across the three ALBI grades. This line therapies, the ALBI grade was predictive of survival
was not replicated in Child B patients where the ALBI with median OS of 17.5 months (95% HR 9.0–26) for
was not prognostic. ALBI 1 (n = 24), 7.5 months for ALBI 2 (95% CI 5.4–9.5)

718 Aliment Pharmacol Ther 2017; 45: 714–722

ª 2017 John Wiley & Sons Ltd

and 1.9 months for ALBI 3 (95% CI 1.4–2.4; log-rank
P < 0.001; Figure 3a). Similarly, patients with Child class
A (n = 135) had better median OS at 10 months (95% CI
7–13) compared to Child B (n = 153) whose median OS
was 2.2 (95% CI 1.6–2.8) (Figure 3b).
In the same subset of 294 patients, we comparatively
tested ALBI grade and Child class for their independent
predictive role in estimating PSOS. We tested ALBI and
Figure 3 | Kaplan–Meier curves describing the overall
survival of candidates for second-line treatment
(n = 294) stratified by albumin–bilirubin (ALBI) grade
(a) and Child class (b).
Aliment Pharmacol Ther 2017; 45: 714–722
ª 2017 John Wiley & Sons Ltd
Albumin-bilirubin grade predicts post-sorafenib survival
Child class in concurrent, multivariable Cox regression
models adjusted for the confounding effect of BCLC
stage at discontinuation and subsequent treatment. Mul-
tivariable analysis of survival confirmed both ALBI and
Child class as independent predictors of PSOS (Table 2).
As shown in Figure 4 ROC curve analysis confirmed
optimal prognostic accuracy of both ALBI and Child
class in predicting 90-days mortality rates, where the
ALBI grade displayed better accuracy with an area under
the ROC curve (AUROC) of 0.69 (95% CI 0.63–0.75)
over Child class 0.66 (0.59–0.72; P < 0.001).
We compared the predictive ability of ALBI and Child
class in estimating PSOS using Harrel’s c-index. In the
whole study population (n = 386), ALBI was superior to
Child class (c-index 0.64 vs. 0.62). ALBI displayed highest
accuracy in patients considered eligible to second-line
therapies (n = 294) (c-index 0.65 vs. 0.63 for Child class),
whereas both scores had significantly lower accuracy in
patients experiencing untreatable disease progression
(n = 92) (c-index 0.54 for ALBI and 0.53 for Child class).
DISCUSSION
The provision of safe and effective systemic anti-cancer
treatment in patients with HCC has been traditionally
hindered by the concurrent presence of liver dysfunction,
which imposes a significant toll on patients’ survival and
ability to tolerate systemic therapies.24 In the drug devel-
opment process, cirrhosis limits enrolment of patients
with HCC to early phase clinical trials due to safety con-
cerns stemming from liver dysfunction, reducing the
detection of early efficacy signals from novel investiga-
tional medicinal products. In later stage trials where
treatment efficacy testing is powered on survival benefit,
liver dysfunction is a major confounder in influencing
the mortality of patients with advanced HCC indepen-
dently from the progression of malignancy.25 Such con-
founding role has affected the clinical development of
sorafenib, where unsurprisingly, despite homogeneity in
screening criteria and treatment schedules, patients
enrolled into the SHARP and Asian Pacific trials had dif-
ferent survival, suggesting that subtle variations in liver
function might have at least in part accounted for the
observed heterogeneity in trial outcomes.6, 7
Accumulating evidence suggests the ALBI grade as a
more accurate predictor of liver functional reserve across
the various stages of HCC compared to standard Child
classification,13, 14 including patients with advanced
HCC.26, 27 In our multi-institutional study including a
large cohort of patients treated with sorafenib, we
demonstrated for the first time the independent
719
D. J. Pinato et al.

Table 2 | Multivariable cox regression analyses evaluating the prognostic role of Child class and ALBI grade in
candidates for second-line systemic therapies (n = 294)

ALBI, Unadjusted ALBI, Adjusted* CTP, Unadjusted CTP, Adjusted*

HR (95% CI) P-value HR (95% CI) P-value HR (95% CI) P-value HR (95% CI) P-value
ALBI
2 vs. 1 1.9 (1.1–3.3) 0.02 1.8 (1.0–3.2) 0.03
3 vs. 1 4.7 (2.7–8.4) <0.001 4.0 (2.2–7.1) <0.001
BCLC
C vs. B 1.8 (1.3–2.4) <0.001 1.9 (1.3–2.5) <0.001
Subsequent therapy
Yes vs. No 0.4 (0.3–0.5) <0.001 0.4 (0.3–0.6) <0.001
Child class
B vs. A 3.0 (2.2–4.0) <0.001 2.4 (1.8–3.4) <0.001
* Coefficients adjusted for post-sorafenib BCLC stage and subsequent therapy status.

in Child A functional class, in whom preservation of an

Figure 4 | The comparison of the albumin–bilirubin
(ALBI) grade and Child class in predicting 90-days
mortality rate using ROC curve analysis (n = 386).
prognostic ability of the ALBI in predicting PSOS. In
unselected patients who failed sorafenib therapy, the
ALBI grade calculated at the moment of permanent sora-
fenib discontinuation was capable of differentiating
patient subgroups with up to a 3-fold difference in med-
ian survival times across the diverse strata.
Interestingly, the prognostic accuracy of the ALBI was
preserved following exclusion of patients who, based on
BCLC stage at sorafenib withdrawal, were unfit for sec-
ond-line therapies. An even stronger and clinically
important stratifying potential of the ALBI grade tran-
spired by restricting our analysis of survival to patients
ALBI grade 1 at treatment discontinuation predicted for
median survival times in excess of 17 months, reducing
to 8 months for ALBI grade 2.
It has been suggested that the improved accuracy of
ALBI, especially within Child A criteria, is secondary to
the modification of pre-defined cut-off points for albu-
min and bilirubin, which might allow clinicians to cap-
ture more subtle differences in the biosynthetic function
of the cirrhotic liver.27–29 Certainly, abandoning scoring
of encephalopathy, ascites or coagulopathy, two of which
require subjective assessments, does not reduce the accu-
racy of the ALBI in predicting survival nor 90-days mor-
tality rates, a landmark survival endpoint that
investigators often use to assess the suitability of patients
to clinical trials. Interestingly, whilst 92% of ALBI grade
1 patients satisfied Child A criteria, such proportion
reduced to 73% in ALBI grade 2 patients and 3% in
ALBI grade 3, suggesting a potentially higher prognostic
heterogeneity in grade 2 patients due to the contribution
of diverse functional subclasses.30
This is not the first study to highlight, amongst other
factors, the key role of deteriorating liver function as a
predictor of survival in patients with advanced HCC
who might be otherwise fit for second-line therapies.
The role of Child class deterioration was reported by
Reig et al., although survival analysis was limited to only
85 patients, who had all discontinued sorafenib due to
disease progression.16 Likewise, despite not having
directly tested Child class, Iavarone and colleagues pro-
vide convincing confirmation of the adverse prognostic
role of a composite panel of individual biomarkers of
impaired liver function including albumin, bilirubin, pro-
thrombin time as well as clinically overt hepatic

720 Aliment Pharmacol Ther 2017; 45: 714–722

ª 2017 John Wiley & Sons Ltd

decompensation as predictors of mortality.10 Based on
our data, the maximal improvement in prognostication
conferred by the ALBI grade is evident in patients who
fulfil Child A criteria, a point which may limit the appli-
cability of this score in patients with intermediate to
advanced liver dysfunction.
In contrast to the previous studies, we prospectively
followed up a larger, multi-institutional cohort of
patients from Europe and Asia, to control for the hetero-
geneity in survival generated by varying aetiological fac-
tors and tumour burden. In addition, our analysis of
survival is also controlled for post-progression treatment
status, an established confounder in estimating OS not
taken into account in previous studies. These method-
ological strengths allow our observation to be controlled
for a number of important sources of systematic bias,
suggesting that the independent prognostic value of the
ALBI is a true finding and generalisable to patients of
diverse geographical origin, ethnicity and aetiology.31
In conclusion, this study validates the ALBI grade as a
prognostic index in HCC patients who permanently dis-
continued sorafenib. The increased accuracy of the ALBI
comes at no additional costs, as it relies on routinely
available, objective biochemical indices. Due to limited
evidence-based treatment options in patients who fail
sorafenib,32 the enhanced discriminative power of the
ALBI in advanced HCC is particularly important for opti-
mal patient stratification, especially in suitable candidates
for clinical trials.33 The accurate definition of patient sub-
populations with prolonged stability in liver functional
reserve might enhance efficacy testing in phase II–III tri-
als by reducing the heterogeneity stemming from con-
comitant liver failure.12 This is particularly important in
trial design, where inaccurate assumptions over predicted
survival times in placebo group as well as the choice of
pre-planned stratification factors has been a problematic
issue in the development of molecularly targeted thera-
pies in the second-line setting.34
Secondly, improved hepatic reserve phenotyping might
help mitigating the risks stemming from systemic toxic-
ity, a nontrivial issue in patients with HCC where treat-
ment discontinuation due to AEs is higher than other
malignancies,35 and has in the past led to the premature
termination of development of sunitinib due to concerns
over excessive mortality in the treatment arm. Although
exploring the relationship between ALBI grade and toxi-
city from systemic treatment was beyond the aims of our
Aliment Pharmacol Ther 2017; 45: 714–722
ª 2017 John Wiley & Sons Ltd
Albumin-bilirubin grade predicts post-sorafenib survival
study, this should be formally evaluated in future
prospective trials.
Although the relatively limited sample size is a poten-
tial limitation to our study, the survival times and clini-
copathological features describing our cohort are similar
to previous studies to suggest our results are generalis-
able to the broader population of patients with HCC
who fail sorafenib therapy.10, 16, 36 Moreover, the choice
of evaluating this score in a large, multi-institutional
patient dataset limits the chances of selection bias.
In the absence of convincingly robust molecular
biomarkers to dissect the biological heterogeneity of
advanced HCC, our study is the first to support the use of
the ALBI as an accurate, objective and reproducible mar-
ker of liver function in HCC. The prognostic relationship
between ALBI and survival in patients should be further
defined in independent studies in a view to facilitate the
full clinical implementation of this promising biomarker.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the
online version of this article:
Table S1. Concordance between Child–Turcotte Pugh
and ALBI-based liver functional classification (n = 377).
Table S2. The prognostic relationship between Child–
Turcotte Pugh and ALBI (n = 342).
AUTHORSHIP
Guarantor of the Article: Dr Rohini Sharma.
Authors contributions: DJP, RS: Study concept and design, DJP, TA,
CY, CW, CS, DB, RT, MK, MP: Acquisition of data, DJP, RS: Anal-
ysis and interpretation of data, DJP, RS: Drafting of the manuscript,
DJP, DB, TA, CY, CW, MP, MB, RT, RR, MK, RS: Critical revision
of the manuscript for important intellectual content, DJP, RR: Statis-
tical analysis, DJP, RS: Obtained funding, MP, MK, RT: Administra-
tive, technical or material support, DJP, RT, MK, MP, RS: Study
supervision. All authors approved the final version of the manu-
script prior to submission.
ACKNOWLEDGEMENTS:
Declaration of personal interests: None.
Declaration of funding interests: DJP is supported by the National
Institute for Health Research (NIHR) as well as grant funding from
the Academy of Medical Sciences and the Imperial NIHR Biomedi-
cal Research Centre (BRC). DB is supported by the Berta-Ottenstein
Programme, Faculty of Medicine, University of Freiburg, Germany.
This project was funded in part by the Academy of Medical Sciences
(AMS, Grant ID SGL013/1021 awarded to DJP). The AMS starter
grant scheme is supported by the Academy of Medical Sciences, The
Wellcome Trust, Medical Research Council, British Heart Founda-
tion, Arthritis Research UK, the Royal College of Physicians and
Diabetes UK.
721
D. J. Pinato et al.
REFERENCES
1. Durand F, Valla D. Assessment of
prognosis of cirrhosis. Semin Liver Dis
2008; 28: 110–22.
2. Worns MA, Weinmann A, Pfingst K,
et al. Safety and efficacy of sorafenib in
patients with advanced hepatocellular
carcinoma in consideration of
concomitant stage of liver cirrhosis. J
Clin Gastroenterol 2009; 43: 489–95.
3. Doyle A, Marsh P, Gill R, et al.
Sorafenib in the treatment of
hepatocellular carcinoma: a multi-centre
real-world study. Scand J Gastroenterol
2016; 51: 979–85.
4. Al-Rajabi R, Patel S, Ketchum NS, et al.
Comparative dosing and efficacy of
sorafenib in hepatocellular cancer
patients with varying liver dysfunction.
J Gastrointest Oncol 2015; 6: 259–67.
5. Federico A, Orditura M, Cotticelli G,
et al. Safety and efficacy of sorafenib in
patients with advanced hepatocellular
carcinoma and Child-Pugh A or B
cirrhosis. Oncol Lett 2015; 9: 1628–32.
6. Llovet JM, Ricci S, Mazzaferro V, et al.
Sorafenib in advanced hepatocellular
carcinoma. N Engl J Med 2008; 359:
378–90.
7. Cheng AL, Kang YK, Chen Z, et al.
Efficacy and safety of sorafenib in
patients in the Asia-Pacific region with
advanced hepatocellular carcinoma: a
phase III randomised, double-blind,
placebo-controlled trial. Lancet Oncol
2009; 10: 25–34.
8. Llovet JM, Bruix J. Molecular targeted
therapies in hepatocellular carcinoma.
Hepatology 2008; 48: 1312–27.
9. Lencioni R, Kudo M, Ye SL, et al.
GIDEON (Global Investigation of
therapeutic DEcisions in hepatocellular
carcinoma and Of its treatment with
sorafeNib): second interim analysis. Int
J Clin Pract 2014; 68: 609–17.
10. Iavarone M, Cabibbo G, Biolato M, et al.
Predictors of survival in patients with
advanced hepatocellular carcinoma who
permanently discontinued sorafenib.
Hepatology 2015; 62(3): 784–91.
11. Pinter M, Sieghart W, Hucke F, et al.
Prognostic factors in patients with
advanced hepatocellular carcinoma
treated with sorafenib. Aliment
Pharmacol Ther 2011; 34: 949–59.
12. Maida M, Iavarone M, Raineri M,
Camma C, Cabibbo G. Second line
systemic therapies for hepatocellular
carcinoma: reasons for the failure.
World J Hepatol 2015; 7: 2053–7.
13. Johnson PJ, Berhane S, Kagebayashi C,
et al. Assessment of liver function in
patients with hepatocellular carcinoma:
722
a new evidence-based approach-the
ALBI grade. J Clin Oncol 2015; 33:
550–8.
14. Hiraoka A, Kumada T, Michitaka K,
et al. Usefulness of albumin-bilirubin
grade for evaluation of prognosis of
2584 Japanese patients with
hepatocellular carcinoma. J
Gastroenterol Hepatol 2016; 31: 1031–6.
15. Chan AW, Kumada T, Toyoda H, et al.
Integration of albumin-bilirubin (ALBI)
score into Barcelona clinic liver cancer
(BCLC) system for hepatocellular
carcinoma. J Gastroenterol Hepatol
2016; 31: 1300–6.
16. Reig M, Rimola J, Torres F, et al.
Postprogression survival of patients
with advanced hepatocellular
carcinoma: rationale for second-line
trial design. Hepatology 2013; 58:
2023–31.
17. Palmer DH, Johnson PJ. Evaluating the
role of treatment-related toxicities in
the challenges facing targeted therapies
for advanced hepatocellular carcinoma.
Cancer Metastasis Rev 2015; 34:
497–509.
18. 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: 908–43.
19. NCI Common Terminology Criteria for
Adverse Events (CTCAE) v.4. 4.03 ed.
Bethesda; 2010.
20. Llovet JM, Bru C, Bruix J. Prognosis of
hepatocellular carcinoma: the BCLC
staging classification. Semin Liver Dis
1999; 19: 329–38.
21. Lencioni R, Llovet JM. Modified
RECIST (mRECIST) assessment for
hepatocellular carcinoma. Semin Liver
Dis 2010; 30: 52–60.
22. Pinato DJ, Stebbing J, Ishizuka M, et al.
A novel and validated prognostic index
in hepatocellular carcinoma: the
inflammation based index (IBI). J
Hepatol 2012; 57: 1013–20.
23. Hanley JA. Receiver operating
characteristic (ROC) methodology: the
state of the art. Crit Rev Diagn Imaging
1989; 29: 307–35.
24. Pinter M, Sieghart W, Graziadei I, et al.
Sorafenib in unresectable hepatocellular
carcinoma from mild to advanced stage
liver cirrhosis. Oncologist 2009; 14: 70–6.
25. Villanueva A, Hernandez-Gea V, Llovet
JM. Medical therapies for hepatocellular
carcinoma: a critical view of the
evidence. Nat Rev Gastroenterol Hepatol
2013; 10: 34–42.
26. Hollebecque A, Cattan S, Romano O,
et al. Safety and efficacy of sorafenib in
hepatocellular carcinoma: the impact of
the Child-Pugh score. Aliment
Pharmacol Ther 2011; 34: 1193–201.
27. Edeline J, Blanc JF, Johnson P, et al. A
multicenter comparison between Child
Pugh and ALBI scores in patients
treated with sorafenib for
hepatocellular carcinoma. Liver Int
2016; 36: 1821–28.
28. Toyoda H, Lai PB, O’Beirne J, et al.
Long-term impact of liver function on
curative therapy for hepatocellular
carcinoma: application of the ALBI
grade. Br J Cancer 2016; 114: 744–50.
29. Chan AW, Chong CC, Mo FK, et al.
Incorporating albumin-bilirubin grade
into the cancer of the liver Italian
program system for hepatocellular
carcinoma. J Gastroenterol Hepatol
2016; doi: 10.1111/jgh.13457
30. D’Amico G, Garcia-Tsao G, Pagliaro L.
Natural history and prognostic
indicators of survival in cirrhosis: a
systematic review of 118 studies. J
Hepatol 2006; 44: 217–31.
31. Kudo M, Lencioni R, Marrero JA, et al.
Regional differences in sorafenib-treated
patients with hepatocellular carcinoma:
GIDEON observational study. Liver Int
2016; 36: 1196–205.
32. Bruix J, Qin S, Merle P, et al.
Regorafenib for patients with
hepatocellular carcinoma who
progressed on sorafenib treatment
(RESORCE): a randomised, double-
blind, placebo-controlled, phase 3 trial.
Lancet 2016; doi: 10.1016/S0140-6736
(16)32453-9 [Epub ahead of print].
33. He AR, Goldenberg AS. Treating
hepatocellular carcinoma progression
following first-line sorafenib:
therapeutic options and clinical
observations. Therap Adv Gastroenterol
2013; 6: 447–58.
34. Llovet JM, Decaens T, Raoul JL, et al.
Brivanib in patients with advanced
hepatocellular carcinoma who were
intolerant to sorafenib or for
whom sorafenib failed: results from the
randomized phase III BRISK-PS study.
J Clin Oncol 2013; 31: 3509–16.
35. Bolos D, Finn RS. Systemic therapy in
HCC: lessons from brivanib. J Hepatol
2014; 61: 947–50.
36. Terashima T, Yamashita T, Takata N,
et al. Post-progression survival and
progression-free survival in patients
with advanced hepatocellular carcinoma
treated by sorafenib. Hepatol Res 2016;
46: 650–6.
Aliment Pharmacol Ther 2017; 45: 714–722
ª 2017 John Wiley & Sons Ltd