This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Cancer Biol Med 2021. doi: 10.20892/j.issn.2095-3941.2020.0207
ORIGINAL ARTICLE
Diagnostic value of 5 serum biomarkers for hepatocellular carcinoma with different epidemiological backgrounds: A large-scale, retrospective study
Dongming Liu1*, Yi Luo2*, Lu Chen1*, Liwei Chen2, Duo Zuo3, Yueguo Li3, Xiaofang Zhang4, Jing Wu5, Qing Xi2, Guangtao Li2, Lisha Qi6, Xiaofen Yue7, Xiehua Zhang8, Zhuoyu Sun9, Ning Zhang10, Tianqiang Song1, Wei Lu1, Hua Guo2
1Department of Hepatobiliary, Liver Cancer Research Center for Prevention and Therapy; 2Department of Tumor Cell Biology; 3Clinical Laboratory, Tianjin Medical University Cancer Institute and Hospital, National Clinical Research Center for Cancer, Key Laboratory of Cancer Prevention and Therapy, Tianjin, Tianjin’s Clinical Research Center for Cancer, Tianjin 300060, China; 4Medical Laboratory, Tianjin Medical University General Hospital, Tianjin 300052, China; 5Clinical Laboratory, Tianjin Third Central Hospital, Tianjin 300170, China; 6Department of Pathology, Tianjin Medical University Cancer Institute and Hospital, National Clinical Research Center for Cancer, Key Laboratory of Cancer Prevention and Therapy, Tianjin, Tianjin’s Clinical Research Center for Cancer, Tianjin 300060, China; 7Department of Tianjin Research Institute of Liver Diseases, Tianjin Second People’s Hospital, Tianjin 300192, China; 8Department of Infectious Diseases, The First Affiliated Hospital of Baotou Medical College, Inner Mongolia University of Science and Technology, Baotou 014010, China; 9Department of Epidemiology and Biostatistics, School of Public Health, Tianjin Medical University, Tianjin 300070, China; 10The Center for Translational Cancer Research, Peking University First Hospital, Beijing 100034, China
ABSTRACT Objective: Hepatocellular carcinoma (HCC) is a lethal global disease that requires an accurate diagnosis. We assessed the potential of
5 serum biomarkers (AFP, AFU, GGT-II, GPC3, and HGF) in the diagnosis of HCC.
Methods: In this retrospective study, we measured the serum levels of each biomarker using ELISAs in 921 participants, including
298 patients with HCC, 154 patients with chronic hepatitis (CH), 122 patients with liver cirrhosis (LC), and 347 healthy controls
from 3 hospitals. Patients negative for hepatitis B surface antigen and hepatitis C antibody (called “NBNC-HCC”) and patients
positive for the above indices (called “HBV-HCC and HCV-HCC”) were enrolled. The selected diagnostic model was constructed
using a training cohort (n = 468), and a validation cohort (n = 453) was used to validate our results. Receiver operating characteristic
analysis was used to evaluate the diagnostic accuracy.
Results: The α-L-fucosidase (AFU)/α-fetoprotein (AFP) combination was best able to distinguish NBNC-HCC [area under the
curve: 0.986 (95% confidence interval: 0.958–0.997), sensitivity: 92.6%, specificity: 98.9%] from healthy controls in the test cohort.
For screening populations at risk of developing HCC (CH and LC), the AFP/AFU combination improved the diagnostic specificity
for early-stage HCC [area under the curve: 0.776 (0.712–0.831), sensitivity: 52.5%, specificity: 91.6% in the test group]. In all-stage
HBV-HCC and HCV-HCC, AFU was also the best candidate biomarker combined with AFP [area under the curve: 0.835 (0.784–
0.877), sensitivity 69.1%, specificity: 87.4% in the test group]. All results were verified in the validation group.
Conclusions: The AFP/AFU combination could be used to identify NBNC-HCC from healthy controls and hepatitis-related HCC
260 Liu et al. Serum biomarkers for HCC with different epidemiological backgrounds
0
200
400
600
CH LC HC
NBNC-HCC
HBV- and HCV-HCC
CH LC HC
NBNC-HCC
HBV- and HCV-HCC
CH LC HC
NBNC-HCC
HBV- and HCV-HCC
CH LC HC
NBNC-HCC
HBV- and HCV-HCC
CH LC HC
NBNC-HCC
HBV- and HCV-HCC
CH LC HC
NBNC-HCC
HBV- and HCV-HCC
CH LC HC
NBNC-HCC
HBV- and HCV-HCC
Test group
Test group
Test group Validation group
Validation group
NS NS
NS
NS
Test group
Test group
A
C
E
G
F
D
B
0
10,000
15,000
5,000
5,000
0
10,000
15,000
25,000
20,000
15,000
10,000
5,000
0
400
300
200
100
0
800
600
400
200
0
0
1
2
3
4
5
AFP
(ng/
mL)
AFP
(ng/
mL)
GPC
3 (n
g/m
L)H
GF
(ng/
mL)
AFU
(mU
/mL)
GG
T-II
(mU
/mL)
AFU
(mU
/mL)
*****
***
***
*****
*** ******
***
******
******
*****
***
*
*** ***
******
******
Figure 1 The median plasma levels of AFP (A), AFU (B), GPC3 (C), GGT-II (D), and HGF (E) in the test cohort and AFP (F) and AFU (G) in the validation cohort. HC, healthy controls; CH, chronic hepatitis; LC, liver cirrhosis; HCC, hepatocellular carcinoma. *P < 0.05; **P < 0.01; ***P < 0.001; P > 0.05 means no significance (NS).
Cancer Biol Med Vol 18, No 1 February 2021 261
The combined AFP/AFU panel showed an improvement in the diagnostic sensitivity for the detection of all-stage and early-stage hepatitis-related HCC
The AUC values of AFP, AFU, GPC3, GGT-II, and HGF in the
all-stage HBV-HCC and HCV-HCC groups were 0.780, 0.752,
0.520, 0.547, and 0.735, respectively (Figure 4A–4E). Because
there was no significance between GGT-II and GPC3 in
detecting all-stage HBV-HCC and HCV-HCC, we chose AFP,
AFU, and HGF in a combination model (Figure 4F–4I). The
diagnostic performance of the serum biomarkers in different
subgroups was further evaluated. Among these combinations,
the AFP/AFU panel outperformed the others and exhibited a
0 0.2 0.4
AUC = 0.79295% CI (0.729–0.845)
AUC = 0.82495% CI (0.764–0.874)
AUC = 0.98995% CI (0.963–0.998)
AUC = 0.98995% CI (0.963–0.998)
AUC = 0.98695% CI (0.958–0.997)
AUC = 0.75995% CI (0.694–0.817)
AUC = 0.98995% CI (0.963–0.998)
AUC = 0.96795% CI (0.932–0.987)
AUC = 0.82595% CI (0.766–0.875)
1-speci�city
1-speci�city
1-speci�city 1-speci�city
Sens
itivi
ty
Sens
itivi
ty
AFP
GGT-II HGF AFP + AFU + GGT-II + GPC3 + HGF
AFP + AFU
AFU GPC3
0.6 0.8 1.0
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0Se
nsiti
vity
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0Se
nsiti
vity
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
A
D
G
E
H I
B C
F
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
AFP + AFU + HGF AFP + AFU + GPC3 + HGF
Figure 2 The receiver operating characteristic curve of AFP (A), AFU (B), GPC3 (C), GGT-II (D), HGF (E), AFP + AFU + GGT-II + GPC3 + HGF (F), AFP + AFU + GPC3 + HGF (G), AFP + AFU + HGF (H), and AFP + AFU (I) in the detection of the NBNC-HCC test group. The sensitivity and specificity represented by the red dots are shown in detail (lower). (A). AFP sensitivity: 59.3% and specificity: 98.9%; (B). AFU sensitivity: 85.2% and specificity: 98.9%; (C). GPC3 sensitivity: 100.0% and specificity: 72.6%; (D). GGT-II sensitivity: 92.6% and specificity: 58.3%; (E). HGF sensi-tivity: 51.9% and specificity: 88.6%; (F). AFP + AFU + GPC3 + GGT-II + HGF sensitivity: 92.6% and specificity: 98.9%; (G). AFP + AFU + GPC3 + HGF sensitivity: 92.6% and specificity: 99.4%; (H). AFP + AFU + HGF sensitivity: 92.6% and specificity: 99.4%; (I). AFP + AFU sensitivity: 92.6% and specificity: 98.9%.
262 Liu et al. Serum biomarkers for HCC with different epidemiological backgrounds
greater diagnostic sensitivity and specificity for the differen-
tiation of all-stage HBV-HCC and HCV-HCC patients from
CH and LC patients [AUC: 0.835 (0.784–0.877), sensitivity:
69.1%, specificity: 87.4%] (Figure 4G). The diagnostic values
of serum AFP and AFU were 42.34 ng/mL and 13.94 mU/mL,
respectively. Regarding early stage HBV-HCC and HCV-HCC,
the AUC values of AFP, AFU, GPC3, GGT-II, and HGF were
0.741, 0.666, 0.517, 0.510, and 0.665, respectively (Figure
5A–5E). We observed similar results in this test cohort with
the all-stage HBV-HCC and HCV-HCC groups. AFP, AFU, and
HGF were selected for the combination model (Figure 5F–5I).
The AFP/AFU combination was also notable for early-stage
HBV-HCC and HCV-HCC [AUC: 0.776 (0.712–0.831), sen-
sitivity: 52.5%, specificity: 91.6%] in the test cohort (Figure
5G). In summary, the AFP/AFU panel improved the diagnos-
tic sensitivity without a loss of specificity in the detection of
all-stage HBV-HCC and HCV-HCC (Table 2: AFP vs. AFP +
AFU: sensitivity 52.8% vs. 69.1%, specificity 93.7% vs. 87.4%)
and early-stage HBV-HCC and HCV-HCC (Table 2: AFP vs.
AFP + AFU, sensitivity: 44.3% vs. 52.5%, specificity: 93.7% vs.
91.6%) among at-risk patients.
In the validation cohort, the concentrations of AFP and
AFU in CH-, LC-, and hepatitis-related HCC patients are
shown in Figure 1F and 1G. The results were similar to those
in the test cohort. The ROC curves of single serum markers
and combined serum markers in the validation group are
shown in Supplementary Figure S6 (the results for all-stage
HBV-HCC and HCV-HCC are shown in A, B, and E; the
results for early-stage HBV-HCC and HCV-HCC are shown
in C, D, and F). Compared with the optimum diagnostic cut-
off values of AFP and AFU for HBV-HCC and HCV-HCC in
the test group, the parameter values in the validation group
for all-stage and early-stage hepatitis-related HCC are sum-
marized in Table 2 [AUC: 0.841 (0.790–0.884), sensitivity:
71.9%, specificity: 86.5% in the validation cohort for all-stage
HBV-HCC and HCV-HCC; AUC: 0.791 (0.728–0.845), sensi-
tivity: 75.4%, specificity: 73.7% in the validation cohort for
early-stage HBV-HCC and HCV-HCC]. The AUC of the AFP/
AFU combination was better than any other single biomarker
(only AFP or AFU) of all-stage (Figure 6A and 6B) and
early-stage (Figure 6C and 6D) hepatitis-related HCC in the
test and validation groups. We also constructed a nomogram
model for the clinical application of these 2 serum markers in
HBV-HCC and HCV-HCC (Figure 6E). For example, if the
AFP and AFU values of an “at-risk person” (such as an indi-
vidual with HBV or HCV) were 60 ng/mL and 25 mU/mL, Tabl
e 1
Resu
lts fo
r the
mea
sure
men
t of s
erum
AFU
, AFP
, or b
oth,
in th
e di
agno
sis
of N
BNC-
HCC
Test
Valid
atio
n
AUC
Sen
sitiv
ity S
peci
ficity
PPV
NPV
Pos
itive
Neg
ativ
e P
AUC
Sen
sitiv
ity S
peci
ficity
PPV
NPV
Pos
itive
Neg
ativ
e P
(95%
CI)
(%
) (
%)
(%
) (
%)
LR
LR
val
ue(9
5% C
I) (
%)
(%
) (
%)
(%
) L
R
LR v
alue
NBN
C-H
CC v
s. H
C (re
sults
for m
easu
rem
ent o
f AFU
, AFP
, or b
oth
in th
e di
agno
sis
of N
BNC-
HCC
)
AFP
0.7
92 (0
.729
–0.8
45)
59.3
% 9
8.9%
88.
9% 9
4.0%
51.
85 0
.41
<0.
001
0.7
07 (0
.639
–0.7
69)
51.9
% 9
7.7%
77.
8% 9
2.8%
22.
25 0
.49
0.0
02
AFU
0.9
67 (0
.932
–0.9
87)
85.2
% 9
8.9%
92.
0% 9
7.7%
74.
54 0
.15
<0.
001
0.9
48 (0
.907
–0.9
74)
74.1
% 9
6.5%
76.
9% 9
6.0%
21.
23 0
.27
<0.
001
AFP
+ A
FU 0
.986
(0.9
58–0
.997
) 92
.6%
98.
9% 9
2.6%
98.
9% 8
1.02
0.0
75 <
0.00
1 0.
969
(0.9
34–0
.988
) 88
.9%
94.
8% 7
2.7%
98.
2% 1
6.99
0.1
2 <
0.00
1
The
diag
nost
ic c
utof
f val
ues
of s
erum
AFP
and
AFU
wer
e 43
.23
ng/m
L an
d 16
.75
mU
/mL,
resp
ectiv
ely.
Cancer Biol Med Vol 18, No 1 February 2021 263
respectively, then based on the nomogram model, the prob-
ability of this individual developing HBV-HCC and HCV-
HCC was nearly 75%.
Overall, the AFP/AFU panel improved the accuracy for
diagnosing all-stage and early-stage hepatitis-related HCC
compared to any other single marker. Moreover, the inclusion
of demographic characteristics assisted in the detection of
disease.
The AFP/AFU combination was effective in predicting the survival of HCC patients
We evaluated the AFU levels in predicting HCC patient prog-
noses based on a KM plotter database. The results showed that
patients with low expression of AFU might have better prog-
noses (Supplementary Figure S7A and B). Thus, we assessed
the value of the AFP/AFU combination in forecasting survival
of HCC patients. The 5-year overall/DFS of HCC patients
with low expression of the AFP/AFU combination was almost
60% and 40%, respectively, while the survival of patients with
high expression was approximately 40% and 25%, respec-
tively (Supplementary Figure S7C and D). Overall, the results
showed that the AFP/AFU combination was effective in pre-
dicting HCC prognosis.
We also verified the results of the KM plotter based on our
IHC data. First, we found that patients with high AFU levels
had worse prognoses (Supplementary Figure S7E and F). In
addition, the IHC results of the AFP/AFU combination in pre-
dicting HCC prognoses were consistent with those in the KM
plotter database (Supplementary Figure S7G and H). Thus,
the AFP and AFU panel was effective in predicting the survival
of HCC patients.
00
Points
AFP (ng/mL)
AFU (mU/mL)
Total points
Probability of HCC
0.2 0.4 0.6
1-specificity
0.001
0
0
0
0 10 20 30 40 50 60 70 80 90 100
20 40 60 80 100 120 140 160 180
5 10 15 20 25 30 ≥35
≥200
20 40 60 80 100 120 140 160 180 200
0.01 0.1 0.4 0.8 0.95 0.999
Nomogram-NBNC-HCC vs. HC
Sens
itivi
ty
Sens
itivi
ty
NBNC-HCC vs. HC (test) NBNC-HCC vs. HC (validation)
0.8 1.0 0 0.2 0.4 0.6
1-specificity
0.8 1.0
AFP + AFU
AFP
AFU
AFP + AFU
AFP
AFU
0.2
0.4
0.6
0.8
1.0
A
C
B
0
0.2
0.4
0.6
0.8
1.0
Figure 3 Diagnostic outcomes and nomograms for the combination of serum AFP and AFU in the diagnosis of NBNC-HCC. (A). Receiver operating characteristic curves (ROCs) for AFU, AFP, or both for all patients with NBNC-HCC vs. HC in the test cohort. (B). ROC curves for AFU, AFP, or both for all patients with NBNC-HCC vs. HC in the validation cohort. (C). Nomogram of the combined AFP/AFU in diagnosing NBNC-HCC.
264 Liu et al. Serum biomarkers for HCC with different epidemiological backgrounds
Discussion
There is a consensus that early diagnosis is the key to improv-
ing the survival of HCC patients4. Several preliminary studies
have suggested that serum biomarkers, including AFP, AFU,
GGT-II, HGF, and GPC3, may be used for the diagnosis of
HCC19-22. However, these markers are not currently included
in routine clinical assessments because of the lack of large-
scale, multicenter clinical investigations.
Over the past 2 decades, infection with hepatitis B virus
(HBV) or hepatitis C virus (HCV) has been associated with
approximately 85% of worldwide HCC32. Due to the promotion
of antiviral therapy, the number of patients with other causes of
HCC (hepatitis B virus surface antigen-negative and hepatitis
0 0.2 0.4
AUC = 0.78095% CI (0.725–0.828)
AUC = 0.54795% CI (0.485–0.608)
AUC = 0.83595% CI (0.784–0.877)
AUC = 0.83695% CI (0.786–0.878)
AUC = 0.76195% CI (0.705–0.811)
AUC = 0.73595% CI (0.678–0.787)
AUC = 0.83695% CI (0.786–0.879)
AUC = 0.75295% CI (0.696–0.803)
AUC = 0.52095% CI (0.458–0.581)
1-speci�city
1-speci�city
1-speci�city 1-speci�city
Sens
itivi
ty
Sens
itivi
ty
AFP
GGT-II HGF AFP + AFU + HGF
AFU + HGF
AFU GPC3
0.6 0.8 1.0
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0Se
nsiti
vity
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0Se
nsiti
vity
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
A
D
G
E
H I
B C
F
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
AFP + HGF AFP + AFU
Figure 4 The receiver operating characteristic curves of AFP (A), AFU (B), GPC3 (C), GGT-II (D), and HGF (E), AFP + AFU + HGF (F), AFP + AFU (G), AFP + HGF (H), AFU + HGF (I) in the detection of all-stage HBV-HCC and HCV-HCC of the test group. The sensitivity and specificity repre-sented by the red dots are shown in detail (lower). (A). AFP sensitivity: 52.8% and specificity: 93.7%; (B). AFU sensitivity: 71.5% and specificity: 67.1%; (C). GPC3 sensitivity: 91.1% and specificity: 25.9%; (D). GGT-II sensitivity: 73.2% and specificity: 38.5%; (E). HGF sensitivity: 61.8% and specificity: 75.5%; (F). AFP + AFU + HGF sensitivity: 65.9% and specificity: 89.5%; (G). AFP + AFU sensitivity: 69.1% and specificity: 87.4%; (H). AFP + HGF sensitivity: 74.8% and specificity: 79.0%; (I). AFU + HGF sensitivity: 63.4% and specificity: 76.9%.
Cancer Biol Med Vol 18, No 1 February 2021 265
C virus antibody-negative or NBNC-HCC) is increasing8,11. In
our study of 401 subjects (202 in the test cohort and 199 in
the validation cohort), the levels of all 5 markers were signifi-
cantly higher in NBNC-HCC patients than in healthy controls.
We therefore further studied the diagnostic capabilities of these
5 markers for NBNC-HCC. It is worth mentioning that the
healthy controls in this study only referred to individuals who
had not been infected with HBV or HCV. The healthy controls
may have suffered from alcohol-related liver disease, nonalco-
holic steatohepatitis, or aflatoxin exposure33. The combination
of AFP and AFU was uniquely associated with the progres-
sion of NBNC-HCC (HC to NBNC-HCC). This combination
showed promising characteristics as a diagnostic marker for
NBNC-HCC. Their diagnostic capability outperformed that of
AUC = 0.74195% CI (0.675–0.800)
AUC = 0.51095% CI (0.440–0.581)
AUC = 0.77695% CI (0.712–0.831)
AUC = 0.77495% CI (0.710–0.829) AUC = 0.666
95% CI (0.597–0.730)
AUC = 0.66595% CI (0.596–0.730)
AUC = 0.77095% CI (0.706–0.826)
AUC = 0.66695% CI (0.597–0.730)
AUC = 0.51795% CI (0.447–0.588)
1-speci�city
1-speci�city
1-speci�city 1-speci�city
Sens
itivi
ty
Sens
itivi
ty
AFP
GGT-II HGF AFP + AFU + HGF
AFU + HGF
AFU GPC3
0 0.2 0.4 0.6 0.8 1.0
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
1-speci�city
0 0.2 0.4 0.6 0.8 1.0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0Se
nsiti
vity
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
A
D
G
E
H I
B C
F
0.2
0
0.4
0.6
0.8
1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
AFP + HGF AFP + AFU
Figure 5 The receiver operating characteristic curves of AFP (A), AFU (B), GPC3 (C), GGT-II (D), and HGF (E), AFP + AFU + HGF (F), AFP + AFU (G), AFP + HGF (H), and AFU + HGF (I) in the detection of early-stage HBV-HCC and HCV-HCC of the test group. The sensitivity and specificity represented by the red dots are shown in detail (lower). (A). AFP sensitivity: 44.3% and specificity: 93.7%; (B). AFU sensitivity: 63.9% and spec-ificity: 67.1%; (C). GPC3 sensitivity: 86.9% and specificity: 28.0%; (D). GGT-II sensitivity: 96.7% and specificity: 11.9%; (E). HGF sensitivity: 50.8% and specificity: 75.5%; (F). AFP + AFU + HGF sensitivity: 52.5% and specificity: 90.2%; (G). AFP + AFU sensitivity: 52.5% and specificity: 91.6%; (H). AFP + HGF sensitivity: 50.8% and specificity: 90.9%; (I). AFU + HGF sensitivity: 34.4% and specificity: 93.0%.
266 Liu et al. Serum biomarkers for HCC with different epidemiological backgrounds
any other serum marker in this study (AUC: 0.986, 95% CI:
0.958–0.997, sensitivity: 92.6%, specificity: 98.9% in the test
specificity: 94.8% in the validation cohort). Considering vari-
ous factors, such as the incidence of NBNC-HCC, our study is
the first large-scale, retrospective analysis of serum biomarkers
in NBNC-HCC patients.
In China, the incidence and mortality of hepatitis-related
HCC is still high34. The HBV-HCC and HCV-HCC patient
median plasma levels of AFP, AFU, and HGF were found to
be significantly higher than those of CH and LC patients.
We showed that the rise of these 3 serum biomarkers may be
related to the occurrence of HBV-HCC and HCV-HCC. Thus,
we paid particular attention to these 3 serum markers for dif-
ferentiating HBV-HCC and HCV-HCC patients from at-risk
(CH and LC) patients. This differentiation has also been the
focus of current research worldwide35. In our study, AFU
showed promising accuracy in identifying HBV-HCC and
HCV-HCC patients from the at-risk population. We found
that the combination of AFP and AFU uniquely reflected the
progression of HBV-HCC and HCV-HCC (CH to LC to HBV-
HCC and HCV-HCC). For all-stage hepatitis-related HCC
vs. CH and LC, the ROC curves showed that the AFP/AFU
combination had an AUC of 0.835 (95% CI: 0.784–0.877), a
sensitivity of 69.1%, and a specificity of 87.4%. Our results
are comparable with other promising markers, especially in
terms of diagnostic sensitivity (e.g., DKK1: 74.8% vs. 69.1%
in all-stage detection)12. Similar results were also shown in
early-stage HBV-HCC and HCV-HCC (AUC: 0.776, 95%
CI: 0.712–0.831, sensitivity: 52.5%, specificity: 91.6%). Most
importantly, the AFP/AFU panel improved the diagnostic sen-
sitivity in the absence of a loss of specificity in the detection
of HBV-HCC and HCV-HCC. Notably, this strategy showed
an advantage for using an AFP/AFU panel. Our findings
are consistent with the results of basic and clinical research
studies23,36,37. AFU was also considered to be a prognostic and
disease recurrence marker and has been shown to be associ-
ated with metastasis and reduced overall survival38.
Zhang et al.39 assessed the diagnostic value for HCC in com-
bination with AFU, AFP, and TK1. They enrolled participants
including 116 patients with HCC, 109 patients with benign
hepatic diseases (such as hepatitis and liver cirrhosis), and 104
normal subjects. The results showed that the AUC was 0.718
for AFU, 0.832 for AFP, 0.773 for TK1, and 0.900 for the com-
bination of these markers. The results were similar to our data
in the detection of HBV-HCC and HCV-HCC (0.780 for AFP, Tabl
e 2
Resu
lts fo
r mea
sure
men
t of s
erum
AFU
, AFP
, or b
oth,
in th
e di
agno
sis
of H
BV-H
CC a
nd H
CV-H
CC
Test
Val
idat
ion
AUC
Sen
sitiv
ity S
peci
ficity
PPV
NPV
Pos
itive
Neg
ativ
e P
AUC
Sen
sitiv
ity S
peci
ficity
PPV
NPV
Pos
itive
Neg
ativ
e P
(95%
CI)
(%
) (
%)
(%
) (
%)
LR
LR
val
ue(9
5% C
I) (
%)
(%
) (
%)
(%
) L
R
LR v
alue
Hep
atiti
s-H
CC v
s. CH
and
LC
(resu
lts fo
r the
mea
sure
men
t of A
FU, A
FP, o
r bot
h in
dia
gnos
is o
f hep
atiti
s-H
CC)
AFP
0.7
80 (0
.725
–0.8
28)
52.8
% 9
3.7%
87.
8% 6
9.8%
8.4
0 0
.50
<0.
001
0.80
9 (0
.755
–0.8
55)
62.8
% 9
0.2%
85.
4% 7
2.7%
6.4
1 0
.41
<0.
001
AFU
0.7
52 (0
.696
–0.8
03)
71.5
% 6
7.1%
65.
2% 7
3.3%
2.1
8 0
.42
<0.
001
0.72
7 (0
.668
–0.7
81)
69.4
% 6
5.4%
64.
6% 7
0.2%
2.0
0 0
.47
<0.
001
AFP
+ A
FU 0
.835
(0.7
84–0
.877
) 69
.1%
87.
4% 8
2.5%
76.
7% 5
.49
0.3
5 <
0.00
1 0.
841
(0.7
90–0
.884
) 71
.9%
86.
5% 8
2.9%
77.
2% 5
.31
0.3
2 <
0.00
1
Early
hep
atiti
s-H
CC v
s. CH
and
LC
(resu
lts fo
r mea
sure
men
t of A
FU, A
FP, o
r bot
h in
dia
gnos
is o
f ear
ly h
epat
itis-
HCC
)
AFP
0.7
41 (0
.675
–0.8
00)
44.3
% 9
3.7%
75.
0% 7
9.8%
7.0
3 0
.59
<0.
001
0.75
8 (0
.693
–0.8
16)
52.2
% 9
0.2%
73.
5% 7
8.4%
5.3
3 0
.53
<0.
001
AFU
0.6
66 (0
.597
–0.7
30)
63.9
% 6
7.1%
45.
3% 8
1.4%
1.9
4 0
.54
<0.
001
0.67
1 (0
.602
–0.7
36)
56.5
% 6
5.4%
45.
9% 7
4.4%
1.6
3 0
.67
<0.
001
AFP
+ A
FU 0
.776
(0.7
12–0
.831
) 52
.5%
91.
6% 7
2.7%
81.
9% 6
.25
0.5
2 <
0.00
1 0.
791
(0.7
28–0
.845
) 75
.4%
73.
7% 5
9.8%
85.
2% 2
.86
0.3
3 <
0.00
1
The
diag
nost
ic c
utof
f val
ues
of s
erum
AFP
and
AFU
wer
e 42
.34
ng/m
L an
d 13
.94
mU
/mL,
resp
ectiv
ely.
Cancer Biol Med Vol 18, No 1 February 2021 267
00 0.2
HBV- and HCV-HCC vs. CH + LC (test) HBV- and HCV-HCC vs. CH + LC (validation)
Early HBV- and HCV-HCC vs. CH + LC (test)
Nomogram HBV- and HCV-HCC vs. CH and LC
Early HBV- and HCV-HCC vs. CH + LC (validation)
AFP + AFU
AFP
AFU
AFP + AFU
AFP
AFU
AFP + AFU
AFP
AFU
AFP + AFU
AFP
AFU
0.4 0.6
1-specificity
Sens
itivi
ty
Sens
itivi
ty
0.8 1.0
0 0.2 0.4 0.6
1-specificity
0.8 1.0 0 0.2 0.4 0.6
1-specificity
0.8 1.0
0 0.2 0.4 0.6
1-specificity
0.8 1.0
0.2
0.4
0.6
0.8
1.0
A
C
E
D
B
0
Sens
itivi
ty
0.2
0.4
Points
AFP (ng/mL)
AFU (mU/mL)
Total points
Probability of HCC0.1
0 20 40 60 80 100 120 140 160 180
0 10 20 30 40 50 60 70 80 90 100
0 20 40 60 80 100 120 140 160 180
0 5 10 15 20 25 30 ≥.35
≥200
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.95
0.6
0.8
1.0
0
Sens
itivi
ty
0.2
0.4
0.6
0.8
1.0
0
0.2
0.4
0.6
0.8
1.0
Figure 6 Diagnostic outcomes and nomogram for the combination of serum AFP and AFU of all-stage and early stage hepatitis-related hepatocellular carcinoma (HCC). (A). Receiver operating characteristic curves (ROCs) for AFU, AFP, or both for all patients with all-stage HBV-HCC and HCV-HCC vs. CH and LC in the test cohort. (B). ROC curves for AFU, AFP, or both for all patients with all-stage HBV-HCC and HCV-HCC vs. CH and LC in the validation cohort. (C). ROC curves for AFU, AFP, or both for all patients with early-stage HBV-HCC and HCV-HCC vs. CH and LC in the test cohort. (D). ROC curves for AFU, AFP, or both for all patients with early-stage HBV-HCC and HCV-HCC vs. CH and LC in the validation cohort. (E). Nomogram of the combined AFP/AFU in diagnosing HBV-HCC and HCV-HCC.
268 Liu et al. Serum biomarkers for HCC with different epidemiological backgrounds
0.752 for AFU, and 0.835 for the combination). In addition,
we showed that the AFP/AFU combination was effective in
detecting NBNC-HCC patients. Zhu et al.40 found that the
AUCs were 0.80, 0.80, and 0.87 for serum AFU, 5′-NT, and
AFP, respectively. The correlation of AFU and AFP was sig-
nificant. However, they did not identify the combination of
these markers. In addition, the number of participants was too
low (36 for HCC and 36 for healthy controls). Xing et al.24
reported that a combination of AFU and AFP (AUC: 0.582)
did not improve the diagnostic efficacy compared with AFP
(AUC: 0.764) alone for HCC patients. They showed that the
majority of HCC patients (85.5%) had chronic HBV and
only 13 patients (6.9%) had chronic HCV, so there were some
NBNC-HCC patients (7.6%) who were enrolled in the HCC
cohort. Based on the etiology of HCC, patients with hepati-
tis do not progress to NBNC-HCC. This part of NBNC-HCC
patients might therefore cause bias in the results. However,
patients with benign disease would not evolve to HBV or
HCV-related HCC. This phenomenon might lead to a low
AUC of AFU and its combination. We enrolled patients with
hepatitis or liver cirrhosis as controls of hepatitis-related HCC
patients. In addition, healthy controls were used for compari-
son with NBNC-HCC patients. Thus, the results regarding the
AUC of the AFP/AFU combination in our study were more
convincing.
Conclusions
To the best of our knowledge, this is the first report show-
ing the potential of AFU in diagnosing NBNC-HCC and
hepatitis-related HCC, based on a study with a large sample
size and independent validation. Wang et al.21 reported that
preoperative serum AFU is a prognostic predictor of HCC
based on survival prognosis data. We showed that AFU was
a promising diagnostic marker for NBNC-HCC and hepa-
titis-related HCC, with a high degree of accuracy and clini-
cally applicable cut-off concentrations; AFU could also serve
as a reliable second-line marker for the detection of HCC.
The AFP/AFU panel had a high degree of accuracy for dif-
ferentiating NBNC-HCC from healthy controls and hepati-
tis-related HCC in patients at risk for developing HCC. The
assays, which are easy to perform and cost effective, can be
translated into a standard protocol for the clinical diagnosis
of HCC, which may identify asymptomatic patients early for
curative treatments. We are currently conducting a prospec-
tive study to confirm the present findings and to determine
the potential utility of measuring AFP/AFU levels to moni-
tor therapeutic responses, and for the prognostic diagnosis
of HCC.
Grant support
This work was supported by the National Natural Science
Foundation of China (Grant Nos. 81972656 and 31671421),
the Key Project of Tianjin Natural Science Foundation (Grant
No. 18JCZDJC35200), the State Key Project on Infectious
Diseases of China (Grant No. 2018ZX10723204), and the
National 135 Major Project of China (2018ZX10302205).
306 9,504.523 83.511 5,806.052 21.382 1.333 Male 49
615 7,589.057 17.519 5,724.224 76.724 0.468 Male 56
82 6,428.845 32.986 494.283 27.011 0.453 Male 36
64 6,373.376 43.353 1,238.855 15.047 0.924 Male 62
37 6,232.097 104.771 818.198 1.708 1.259 Male 69
16 6,214.831 39.433 875.201 17.254 1.213 Male 45
273 6,209.024 20.951 744.541 19.238 1.223 Male 55
178 6,136.615 159.149 3,136.487 39.649 0.663 Male 45
187 6,071.655 273.419 4,300.544 5.844 0.928 Male 55
397 5,999.737 21.141 5,464.136 2.699 0.302 Male 43
507 5,575.943 66.806 4,566.809 76.817 0.725 Male 48
370 5,285.960 80.587 4,093.100 2.303 0.792 Male 63
Cancer Biol Med Vol 18, No 1 February 2021 27
Number AFP (ng/mL) AFU (mU/mL) GGT2 (mU/mL) GPC3 (ng/mL) HGF (ng/mL) Gender Age
116 5,149.800 48.661 1,802.633 2.313 0.460 Male 53
385 5,111.872 21.484 2,932.634 7.726 0.788 Male 67
382 5,069.321 13.204 766.521 1.535 0.295 Male 65
180 4,713.706 703.627 2,962.245 116.129 1.832 Male 46
475 4,499.951 6.427 965.083 8.211 0.538 Male 31
169 3,518.442 17.517 857.541 5.701 0.667 Male 59
100 3,446.153 39.274 666.546 15.214 1.119 Male 59
388 2,610.837 19.865 10,162.328 8.240 0.738 Male 56
489 2,296.835 11.958 708.857 1.440 0.619 Male 51
96 2,195.566 26.766 342.188 13.209 0.843 Male 57
639 1,865.760 13.756 674.008 29.225 0.362 Male 27
577 1,837.225 20.401 13,565.523 13.708 0.526 Male 50
78 1,661.705 10.208 259.838 24.279 0.415 Male 53
246 1,553.839 18.211 2,735.734 3.725 0.585 Male 64
446 1,372.806 10.520 478.522 1.986 0.365 Male 57
75 1,063.240 32.939 830.405 4.104 0.710 Male 45
461 733.012 12.567 12,522.107 14.483 0.479 Male 48
147 550.388 36.054 1,346.208 18.044 0.906 Male 57
547 541.150 10.316 855.376 9.672 0.301 Male 34
176 519.243 53.613 569.566 14.721 0.403 Male 53
576 477.239 19.977 1,929.334 31.040 0.701 Male 50
456 454.474 13.065 1,013.741 0.000 0.464 Male 44
519 415.595 12.476 3,443.506 7.593 0.483 Male 51
317 404.954 20.603 767.306 21.466 0.945 Male 46
565 359.038 10.235 839.661 144.788 0.338 Male 58
66 305.264 41.759 1,289.495 2.809 1.075 Male 68
392 267.832 30.417 6,689.392 8.830 1.384 Male 58
201 260.572 185.869 995.518 60.244 0.566 Male 53
12 247.384 11.872 369.806 51.479 0.470 Male 59
89 198.840 50.325 593.439 45.623 2.258 Male 44
184 192.441 551.973 6,795.512 11.740 0.825 Male 56
76 185.720 11.915 144.717 36.082 0.388 Male 58
53 178.714 67.456 499.480 2.450 2.363 Male 45
260 164.371 33.079 881.132 23.490 0.902 Male 57
196 152.945 70.806 0.000 8.925 0.612 Male 65
Table S5a (continued)
28 Liu et al. Serum biomarkers for HCC with different epidemiological backgrounds
Number AFP (ng/mL) AFU (mU/mL) GGT2 (mU/mL) GPC3 (ng/mL) HGF (ng/mL) Gender Age
274 147.021 32.624 4,467.932 3.725 0.570 Male 44
134 136.715 34.583 1,371.196 17.293 1.120 Male 62
573 136.375 7.276 2,349.475 34.995 0.548 Male 48
310 133.141 21.734 737.476 1.321 0.772 Male 48
593 95.172 123.940 2,512.192 48.343 0.575 Male 56
148 80.199 24.664 804.114 11.727 1.180 Male 62
653 67.094 7.128 7,067.529 5.339 0.952 Male 52
360 45.236 23.070 474.499 0.508 0.432 Male 56
261 32.513 22.376 4,726.772 12.362 0.657 Male 53
77 31.877 50.151 256.477 36.130 1.551 Male 55
81 25.188 98.377 827.044 33.375 2.118 Male 44
74 25.188 17.196 356.473 3.477 0.498 Male 51
450 10.025 16.964 610.652 43.078 0.695 Male 45
585 9.957 6.734 673.425 18.518 0.262 Male 52
561 8.385 14.289 645.260 9.350 0.552 Male 52
126 6.352 59.547 6,736.497 48.009 1.509 Male 50
41 5.668 16.019 378.950 4.499 0.894 Male 48
165 4.845 10.904 539.545 14.631 0.761 Male 56
534 4.392 4.815 651.662 4.226 0.298 Male 57
425 4.019 15.056 477.629 0.000 0.359 Male 36
118 2.584 17.254 805.625 10.396 0.820 Male 54
594 0.000 24.314 1,531.477 18.154 3.502 Male 50
617 0.000 316.716 15,108.376 30.733 1.689 Male 72
539 0.000 74.438 2,992.270 144.788 1.471 Male 54
88 0.000 21.376 473.276 8.072 1.009 Male 48
447 0.000 56.567 911.517 4.813 0.963 Male 66
172 0.000 206.436 649.783 62.461 0.864 Male 63
614 0.000 28.599 875.912 10.266 0.844 Male 54
171 0.000 20.598 652.352 37.079 0.795 Male 49
30 0.000 18.778 766.308 0.000 0.655 Male 55
455 0.000 38.916 867.771 0.751 0.624 Male 57
599 0.000 17.391 761.693 8.292 0.623 Male 70
499 0.000 18.458 895.447 65.053 0.601 Male 62
162 0.000 11.712 506.271 8.460 0.544 Male 49
108 0.000 17.686 853.619 2.725 0.506 Male 59
Table S5a (continued)
Cancer Biol Med Vol 18, No 1 February 2021 29
Number AFP (ng/mL) AFU (mU/mL) GGT2 (mU/mL) GPC3 (ng/mL) HGF (ng/mL) Gender Age
163 0.000 11.130 289.584 1.394 0.501 Male 58
538 0.000 6.710 425.248 1.602 0.444 Male 63
396 0.000 18.430 846.344 2.960 0.405 Male 69
383 0.000 3.670 600.391 5.825 0.363 Male 59
564 0.000 6.608 382.759 34.513 0.305 Male 66
658 0.000 4.340 899.117 0.000 0.266 Male 64
605 0.000 8.119 737.756 2.249 0.265 Male 52
20 0.000 274.126 3,216.747 3.439 1.822 Male 69
44 123.779 121.444 300.813 4.146 0.890 Male 76
50 439.658 77.904 633.310 0.153 1.159 Male 55
130 70.403 47.763 4,285.802 12.187 1.028 Male 55
168 0.000 45.117 931.627 4.128 1.389 Male 61
611 0.000 21.927 3,670.746 17.888 0.548 Male 74
151 4,520.299 18.982 2,603.688 15.406 0.475 Male 73
107 0.000 15.188 722.958 5.798 0.528 Male 53
619 7,322.315 12.888 1,191.285 20.746 0.284 Male 61
Table S5a (continued)
30 Liu et al. Serum biomarkers for HCC with different epidemiological backgrounds
996 subjects enrolled in Tianjin Medical University CancerInstitute and Hospital, Tianjin Third Central Hospital andTianjin Medical University General Hospital from July 2012to April 2014
75 excludedPrimary liver cancer other than HCC (27): ICC (23)and HCC-CC (4)Metastatic liver cancer (2)Other liver tumor such as sarcoma and adenoma,etc (11)Clinical data not available (35): HCC (34) and HBV(1)
921 subjects enrolled (all with values of AFP, AFU,GGT-II, GPC3 and HGF)
NBNC-HCC group Hepatitis-HCC group
HC NBNC-HCC CH LC Hepatitis-HCC
202 test cohort27 HCC175 HC
199 validation cohort27 HCC172 HC
266 test cohort
•
• •
•
• • •
• • • •
• •
• 82 CH61 LC
254 validation cohort121 HCC72 CH61 LC
202 ELISA testing at Tianjin MedicalUniversity Cancer Institute andHospital and the performance of thefive serum markers assessed by ROCanalysis
266 ELISA testing at Tianjin MedicalUniversity Cancer Institute andHospital and the performance of thefive serum markers assessed by ROCanalysis
Figure S1 The study design. HC, healthy controls; CH, chronic hepatitis; LC, liver cirrhosis; HCC, hepatocellular carcinoma.
Cancer Biol Med Vol 18, No 1 February 2021 31
0 0.2 0.4
AUC = 0.98695% CI (0.958–0.997)
AUC = 0.91395% CI (0.866–0.968)
AUC = 0.98895% CI (0.962–0.998)
1-speci�city
Sens
itivi
ty
AFP + AFU + GGT-II + GPC3
AFP + GGT-II + GPC3 + HGF
AFP + AFU + GGT-II + HGF
0.6 0.8 1.0
0 0.2 0.41-speci�city
0.6 0.8 1.0
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0A
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
C
AUC = 0.98395% CI (0.954–0.996)
AFU + GGT-II + GPC3 + HGF
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0
D
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0B
Figure S2 The 4 arrangements of serum biomarkers in the combination model. The sensitivity and specificity represented by the red dots were shown in detail below. (A). AFP + AFU + GGT-II + GPC3 sensitivity: 92.6% and specificity: 99.4%; (B). AFP + AFU + GGT-II + HGF sensi-tivity: 92.6% and specificity: 98.3%; (C). AFP + GPC3 + GGT-II + HGF sensitivity, 77.8% and specificity: 94.9%; (D). AFU + GPC3 + GGT-II + HGF sensitivity; 88.9% and specificity: 100.0%.
32 Liu et al. Serum biomarkers for HCC with different epidemiological backgrounds
0 0.2 0.4
AUC = 0.98695% CI (0.959–0.997)
AUC = 0.83395% CI (0.774–0.881)
AUC = 0.91595% CI (0.868–0.950)
AUC = 0.98695% CI (0.959–0.997)
1-speci�city
Sens
itivi
ty
AFP + AFU + GGT-II
AFP + GGT-II + GPC3 AFP + GGT-II + HGF
AFP + AFU + GPC3
0.6 0.8 1.0
0 0.2 0.4
1-speci�city
0.6 0.8 1.0 0 0.2 0.4
1-speci�city
0.6 0.8 1.0
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0A
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0D
AUC = 0.97595% CI (0.943–0.992)
AFU + GGT-II + GPC3
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0G
AUC = 0.98395% CI (0.955–0.996)
AFU + GGT-II + HGF
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0H
AUC = 0.97495% CI (0.942–0.991)
AFU + GPC3 + HGF
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0I
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0E
AUC = 0.91495% CI (0.867–0.949)
AFP + GPC3 + HGF
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0F
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0B
AUC = 0.86895% CI (0.814–0.912)
GGT-II + GPC3 + HGF
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0C
Figure S3 The 3 arrangements of serum biomarkers in the combination model. The sensitivity and specificity represented by the red dots are shown in detail below. (A). AFP + AFU + GGT-II sensitivity: 92.6% and specificity: 100.0%; (B). AFP + AFU + GPC3 sensitivity: 92.6% and specificity: 98.9%; (C). GGT-II + GPC3 + HGF sensitivity: 74.1% and specificity: 89.1%; (D). AFP + GGT-II + GPC3 sensitivity: 63.0% and specificity: 98.3%; (E). AFP + GGT-II + HGF sensitivity: 77.8% and specificity: 94.9%; (F). AFP + GPC3 + HGF sensitivity: 77.8% and specificity: 95.4%; (G). AFU + GGT-II + GPC3 sensitivity: 88.9% and specificity: 100.0%; (H). AFU + GGT-II + HGF sensitivity: 88.9% and specificity: 98.9%; (I). AFU + GPC3 + HGF sensitivity: 85.2% and specificity: 97.7%.
Cancer Biol Med Vol 18, No 1 February 2021 33
0 0.2 0.4
AUC = 0.82695% CI (0.767–0.876)
AUC = 0.91695% CI (0.868–0.950)
AUC = 0.97595% CI (0.943–0.992)
AUC = 0.83895% CI (0.780–0.886)
1-speci�city
Sens
itivi
ty
AFP + GGT-II
AFP + HGF AFU + GGT-II
AFP + GPC3
0.6 0.8 1.0
0 0.2 0.4
1-speci�city
0.6 0.8 1.0 0 0.2 0.4
1-speci�city
0.6 0.8 1.0
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0A
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0D
AUC = 0.97495% CI (0.942–0.991)
AFU + HGF
0 0.2 0.41-speci�city
0.6 0.8 1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0G
AUC = 0.83695% CI (0.777–0.884)
GGT-II + GPC3
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0H
AUC = 0.86895% CI (0.814–0.912)
GGT-II + HGF
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0I
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0E
AUC = 0.96695% CI (0.931–0.986)
AFU + GPC3
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0F
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0B
AUC = 0.76595% CI (0.700–0.822)
GPC3 + HGF
0 0.2 0.4
1-speci�city
0.6 0.8 1.0
Sens
itivi
ty
0.2
0
0.4
0.6
0.8
1.0C
Figure S4 The 2 arrangements of serum biomarkers in the combination model. The sensitivity and specificity represented by the red dots are shown in detail below. (A). AFP + GGT-II sensitivity: 63.0% and specificity: 98.3%; (B). AFP + GPC3 sensitivity: 59.3% and specificity: 98.3%; (C). GPC3 + HGF sensitivity: 59.3% and specificity: 83.4%; (D). AFP + HGF sensitivity: 77.8% and specificity: 94.9%; (E). AFU + GGT-II sensitivity: 88.9% and specificity: 100.0%; (F). AFU + GPC3 sensitivity: 85.2% and specificity: 98.9%; (G). AFU + HGF sensitivity: 88.9% and specificity: 94.9%; (H). GGT-II + GPC3 sensitivity: 81.5% and specificity: 72.5%; (I). GGT-II + HGF sensitivity: 74.1% and specificity: 87.4%.
34 Liu et al. Serum biomarkers for HCC with different epidemiological backgrounds
0 0.2 0.4
AUC = 0.70795% CI (0.639–0.769)
AUC = 0.96995% CI (0.934–0.988)
1-speci�city
Sens
itivi
ty
AFP
0.6 0.8 1.0
0 0.2 0.41-speci�city
0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0A
Sens
itivi
ty
0.2
0
0.4
0.6
AFP + AFU
0.8
1.0C
0 0.2 0.4
AUC = 0.94895% CI (0.907–0.974)
1-speci�city
Sens
itivi
ty
AFU
0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0B
Figure S5 The receiver operating characteristic curves of AFP (A) AFU (B) and AFP + AFU (C) in the detection of the NBNC-HCC validation group. The sensitivity and specificity represented by the red dots are shown in detail below. (A). AFP sensitivity: 51.9% and specificity: 97.7%; (B). AFU sensitivity: 74.1% and specificity: 96.5%; (C). AFP + AFU sensitivity: 88.9% and specificity: 94.8%.
Cancer Biol Med Vol 18, No 1 February 2021 35
0 0.2 0.4
AUC = 0.80995% CI (0.755–0.855)
1-speci�city
Sens
itivi
ty
AFP
0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0
A
0 0.2 0.4
AUC = 0.75895% CI (0.693–0.816)
1-speci�city
Sens
itivi
ty
AFP
0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0C
0 0.2 0.4
AUC = 0.84195% CI (0.790–0.884)
1-speci�city
Sens
itivi
ty
AFP + AFU
0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0E
0 0.2 0.4
AUC = 0.79195% CI (0.728–0.845)
1-speci�city
Sens
itivi
ty
AFP + AFU
0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0F
0 0.2 0.4
AUC = 0.67195% CI (0.602–0.736)
1-speci�city
Sens
itivi
ty
AFU
0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0D
0 0.2 0.4
AUC = 0.72795% CI (0.668–0.781)
1-speci�city
Sens
itivi
ty
AFU
0.6 0.8 1.0
0.2
0
0.4
0.6
0.8
1.0
B
Figure S6 The serum biomarkers AFP, AFU, and combination model of all-stage and early-stage hepatitis-hepatocellular carcinoma (HCC) in the validation cohort. (A), (B), and (E) show the AFP, AFU, and the combination model of all-stage hepatitis-HCC; (C), (D), and (F) show AFP, AFU, and the combination model of early-stage hepatitis-HCC. The sensitivity and specificity represented by the red dots are shown in detail below. (A). AFP sensitivity: 62.8% and specificity: 90.2%; (B). AFU sensitivity: 69.4% and specificity: 65.4%; (C). AFP sensitivity: 52.2% and speci-ficity: 90.2%; (D). AFU sensitivity: 56.5% and specificity: 65.4%; (E). AFP + AFU sensitivity: 71.9% and specificity: 86.5%; (F). AFP + AFU sensitivity: 75.4% and specificity: 73.7%.
36 Liu et al. Serum biomarkers for HCC with different epidemiological backgrounds
B C DTCGA datasets-diseasefree survival TCGA datasets-diseasefree survival
LowHigh
AFU expressionLowHigh
AFU expression
LowHigh
LowHigh
AFP/AFU expressionLowHigh
AFP/AFU expressionLowHigh
AFP/AFUexpression
LowHigh
AFP/AFU expressionLowHigh
60Time (month)
Surv
ival
rate
Number at risk
Number at risk
Number at risk Number at riskLow 265 208 245142 83 134 58 29 13 4 064 33 37 17 616 5 1
99 108 119 48 26 13 6 2 140 22 10 3 121
1020 9 3 1 0High
Low 78 73 52 35 6104 63
4932
3522
2814
41
7810437 21 5
00High
Number at riskLowHigh
9442
6227
3917
65
00
11270
00
Number at riskLowHigh
5823
4215
3210
41
00
11270
Number at riskLowHigh
LowHigh
LowHigh
80 100
HR = 1.78 (1.24–2.56)
logrank P = 0.0014
HR = 2.319 (1.603–3.355)
logrank P < 0.0001
HR = 2.309 (1.630–3.271)
logrank P < 0.0001
HR = 1.609 (1.131–2.289)
logrank P = 0.007
HR = 1.831 (1.309–2.562)
logrank P < 0.0001
HR = 1.95 (1.40–2.73)
logrank P = 6.6e–05
HR = 1.66 (1.17–2.36)
logrank P = 0.004
HR = 1.55 (1.12–2.16)
logrank P = 0.0083
120
0 20 40 60Time (month)
80 100 0 20 40 60Time (month)
80 100 0 20 40 60Time (month)
80 100 0 20 40 60Time (month)
80 100
0 20 40 60Time (month) Time (month)
80 100 120 0 20 40 60 80 100 120Time (month)
0 20 40 60 80 100 120
0.2
0.4
0.6
0.8
1.0
0.0
Surv
ival
rate
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
Number at risk176 70 31 11 3 1 0140 35 16 9 4 2 1
LowHigh
Figure S7 The AFU and AFP/AFU combination model for predicting hepatocellular carcinoma (HCC) prognosis. (A), (B) The survival curves [overall survival (OS) and disease-free survival (DFS)] of HCC patients with different expressions of AFU based on The Cancer Genome Atlas (TCGA) database. (C), (D) The survival curves (OS and DFS) of HCC patients with different expressions of AFP/AFU based on TCGA database. (E), (F) The survival curves (OS and DFS) of HCC patients with different expressions of AFU based on our immunohistochemistry (IHC) data. (G), (H) The survival curves (OS and DFS) of HCC patients with different expressions of AFP/AFU based on our IHC data.