of 13
7/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
1/13
DOI:10.1093/jnci/djs372
JNCI | Articles 1599
The Author 2012. Published by Oxford University Press. All rights reserved.
For Permissions, please e-mail: [email protected].
jnci.oxfordjournals.org
Article
Hpaoua canoma rsk Pdon Mod fo h Gna
Popuaon: th Pdv Pow of tansamnass
Chi-Pang Wen*, Jie Lin*, Yi Chen Yang, Min Kuang Tsai, Chwen Keng Tsao, Carol Etzel, Maosheng Huang, Chung Yi Hsu,
Yuanqing Ye, Lopa Mishra, Ernest Hawk, Xifeng Wu
Manuscript received March 01, 2012; revised July 26, 2012; accepted July 27, 2012
*Authors contributed equally to this work.
Correspondence to: Xifeng Wu, MD, PhD, Department of Epidemiology, The University of Texas MD Anderson Cancer Center, 1155 Herman P. Pressler
Blvd, Unit 1340, Houston, TX 77030 (e-mail: [email protected]).
Background Risk prediction models for hepatocellular carcinoma are available for individuals with chronic hepatitis B virus
(HBV) and hepatitis C virus (HCV) infections who are at high risk but not for the general population with average
or unknown risk. We developed five simple risk prediction models based on clinically available data from the
general population.
Methods A prospective cohort of 428 584 subjects from a private health screening firm in Taiwan was divided into two
subgroupsone with known HCV test results (n = 130 533 subjects) and the other without (n = 298 051 subjects).
A total of 1668 incident hepatocellular carcinomas occurred during an average follow-up of 8.5 years. Model
inputs included age, sex, health historyrelated variables; HBV or HCV infectionrelated variables; serum levels of
alanine transaminase (ALT), aspartate transaminase (AST), and alfa-fetoprotein (AFP), as well as other variables
of routine blood panels for liver function. Cox proportional hazards regression method was used to identify risk
predictors of hepatocellular carcinoma. Receiver operating characteristic curves were used to assess discrimina-
tory accuracy of the models. Models were internally validated. All statistical tests were two-sided.
Results Age, sex, health history, HBV and HCV status, and serum ALT, AST, AFP levels were statistically significant inde-
pendent predictors of hepatocellular carcinoma risk (all P < .05). Use of serum transaminases only in a model
showed a higher discrimination compared with HBV or HCV only (for transaminases, area under the curve
[AUC] = 0.912, 95% confidence interval [CI] = 0.909 to 0.915; for HBV, AUC = 0.840, 95% CI = 0.833 to 0.848; and for
HCV, AUC = 0.841, 95% CI = 0.834 to 0.847). Adding HBV and HCV data to the transaminase-only model improved
the discrimination (AUC = 0.933, 95% CI = 0.929 to 0.949). Internal validation showed high discriminatory accuracy
and calibration of these models.
Conclusion Models with transaminase data were best able to predict hepatocellular carcinoma risk even among subjects with
unknown or HBV- or HCV-negative infection status.
J Natl Cancer Inst 2012; 104:15991611
Chronic hepatitis B virus (HBV) and hepatitis C virus (HCV)
infections are three to five times more common than HIV infection
and AIDS in the United States, pacing those infected with HBVor HCV at increased risk for hepatoceuar carcinoma, cirrhosis,
and death (1). However, unike HIV infection and AIDS, a recent
Institute of Medicine (IOM) report noted that most of the five
miion Americans with HBV or HCV infections are unaware
of their risks unti they deveop symptoms of hepatoceuar
carcinoma or cirrhosis (2). Many of the 150 000 deaths expected
in the next 10 years coud be prevented if physicians and the
pubic were better educated about eary recognition of these
conditions.
Athough HBV or HCV carriers are at increased risk of hepa-
toceuar carcinoma, the cancer aso occurs among noncarriers
of these viruses (3,4). In this aspect, cinicians wi have difficuty
assessing this ow-risk popuation. To start checking for carrier
status, HBV or HCV testing wi require extra effort and extracost; however, recommendations for universa screening have been
suggested (5). Even when HBV or HCV testing is performed and
found to be positive, the majority of carriers do not take action
to reduce their risk (2). This is party because of the fact that the
reationship between hepatitis carrier status and hepatoceuar car-
cinoma risk, specific to the individua, is not readiy avaiabe to
the doctors. As a resut, the test information, whether positive or
negative for HBV or HCV, is often wasted. As not knowing ones
risk is a major barrier for taking action (2), much of the recent
progress made on the treatment of HBV or HCV cannot be fuy
utiized (6).
mailto:[email protected]?subject=http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-mailto:[email protected]?subject=7/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
2/13
Vol. 104, Issue 20 | October 17, 20121600 Articles | JNCI
Avaiabe prediction modes for hepatoceuar carcinoma are
imited to individuas at eevated risk who are carrying HBV (79)
or HCV (10,11). Many of these modes aso need additiona infor-
mation from cinica workups, such as presence of HBV DNA
in the bood (7,8), and this has further imited the appicabiity
of the prediction mode because such measures are usuay not
readiy avaiabe in cinica settings. As much of the pubic is una-
ware of their risk profie, the need for a new average-risk mode
is obvious. Finay, assessing risk in the genera pubic withoutmass screening for HBV or HCV is another chaenge, because
conducting such screenings has not been proven effective scien-
tificay (5). Hepatoceuar carcinoma has a high mortaity rate,
and possibe interventions, such as interferon therapy or ifestye
changes, are avaiabe to reduce the mortaity or ater the course
of the disease, provided individuas at high risk can be identified.
A simpe, easy-to-administer risk prediction mode based on com-
mony avaiabe data at heath checkups woud be of great vaue.
Taking advantage of a medica screening program invoving
neary haf a miion heathy individuas in Taiwan with foow-up
data (12,13), we deveoped a prediction mode for hepatoceuar
carcinoma based on data routiney coected in a typica office visit.The intent of the prediction mode is to provide a simpe, efficient,
and widey avaiabe too to identify and quantify cancer risk in the
average-risk popuation.
Mhods
Study Population and Data Collection
The study popuation was obtained from a standard medica
screening program conducted by the MJ Heath Management
Institution (MJ). From 1994 to 2008, a tota of 428 584 subjects,
free of cancer at baseine, were recruited. In the MJ cohort, because
tests for HCV infection were performed at an extra cost to mem-
bers, ony a subset of 130 533 participants has data on HCV status.Given the importance of HCV as a risk factor for hepatoceuar
carcinoma, we divided the cohort into two subcohorts in order to
provide more accurate risk prediction estimates: one cohort had
the HCV test (n = 130 533 subjects), and the other had no HCV
test (n = 298 051 subjects). Participants in the MJ cohort were aged
20 years or oder.
A participants competed a sef-administered questionnaire
covering demographic characteristics and heath history, incud-
ing ifestye and medica history (such as diabetes, hypertension,
stroke, heart diseases). Subjects who sef-reported having been
diagnosed with diabetes or currenty taking diabetes medica-
tion were defined as having diabetes. A subjects went throughtesting for anthropometric measurements (eg, height, weight,
waist circumference, hip circumference, body fat percentage,
etc.), bood pressure, puse rate, respiration rate, and chest cir-
cumference. Overnight fasting bood was anayzed for a stand-
ard pane, incuding hemogram, bood sugar tests, iver function
tests, rena function tests, bood ipid tests, thyroid function tests,
bood grouping, the presence of HBV surface antigen in bood
(HBsAg), and the presence of HCV antibody in bood (offered
to a subgroup of members with additiona cost). Individuas who
tested positive for HBsAg are referred to as HBV+ subjects and
those tested positive for HCV antibody are referred to as HCV+
subjects, and individuas who tested negative were referred to as
HBV and HCV subjects, respectivey. Smoking was cas-
sified by the number of pack-years (ie, daiy cigarette quantity
duration in years) among ever-smokers. Acoho consumption was
cassified into reguar drinkers (those who consumed 2 drinks/
day on 3 days/week) and occasiona drinkers (those who con-
sumed
7/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
3/13
JNCI | Articles 1601jnci.oxfordjournals.org
Four modes were deveoped in the subcohort without an
HCV test (Tabe 1), and a fifth mode was added in the subcohort
with an HCV test (Tabe 2). Modeing started with heath history
(mode 1: age, sex, pack-year of smoking, acoho drinking, physi-
ca activity, and diabetes) or transaminase ony (mode 2: age, sex,
AST, and ALT), foowed by combination of these variabes into
a joint transaminase and heath history mode (mode 3: age, sex,
pack-year of smoking, acoho drinking, physica activity, diabetes,
AST, and ALT). Finay, mode 3 was extended by adding HBV test
resuts and AFP (mode 4: age, sex, pack-year of smoking, aco-
ho drinking, physica activity, diabetes, AST, ALT, AFP, and HBV)
and further extended by adding HBV and HCV test resuts and
AFP (mode 5: age, sex, pack-year of smoking, acoho drinking,
physica activity, diabetes, AST, ALT, AFP, HBV, and HCV) vari-
abes. HBV and HCV variabes were dichotomized as positive or
negative as described above. In seecting the cutoff points for AST
and ALT, we set the reference group at different starting points
(eg,
7/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
4/13
Vol. 104, Issue 20 | October 17, 20121602 Articles | JNCI
each additiona 5 IU/L at a time through mutipe iterations. We
found that the reference at 25 IU/L was best to differentiate risk
groups (ie, there was a substantia increase in HRs when exceeding
25 IU/L in a scenarios). It is to be noted that the 25 IU/L cutoff
point is much ower than the upper imit of the norma reference
(ULN), usuay set around 40 IU/L (17).
Mode goodness of fit was assessed in terms of discriminatory
accuracy and caibration in an interna vaidation. Discriminatory
accuracy for predicting the deveopment of hepatoceuar carci-noma within 10 years was assessed by constructing time-dependent
receiver operating characteristic curves for censored surviva data
(18) and cacuating the area under the curve (AUC). The adequacy
of each fitted mode was aso evauated by cacuating the concord-
ance index (C-index), which aso measures the mode discrimina-
tory accuracy, using the training and vaidation sets. Simiar to
AUC, we cacuated C-index based on a 10-year prediction. A tota
of four datasets were used in cacuating the C-index: one training
set and one vaidation set for those with an HCV test, and one
training set and one vaidation set for those without an HCV test.
We assessed interna caibration of the modes by determining the
extent of agreement between predicted and observed events in10 years (ie, caibration) (19) and then created a cross-vaidated ca-
ibration pot. We used the whoe study popuation to perform the
10-fod cross-vaidated caibration for different modes, in which
the study popuation was randomy divided into 10 equa sub-
sets with nine subsets as training set and one subset as testing set.
Cross-vaidated predicted probabiity was cacuated in each decie.
The 5- and 10-year absoute risks were cacuated from baseine
probabiity and reative risk profie from the Cox proportiona haz-
ard regression mode, using the standard equation for surviva data
with censored observations (20):
F t S tb X b M j j j jj
p
j
p
( ) ( )exp
= [ ] ( )( )==1 110
whereF(t) denotes the probabiity of deveoping cancer in tyears;
S0(t) is the baseine surviva function; bjis the regression coefficient
for thejth variabe (Xj);Mj denotes the mean eve ofXj;p is the
number of variabes.
We derived risk scores for each statisticay significant predictor
based on regression coefficients in the Cox proportiona hazards
regression mode foowing the reported procedures (21). In a
modes, reference eve of a particuar risk factor received a risk
score of zero. For a particuar risk factor, risk score was assigned
as integer points to each risk eve and cacuated as a weighted
distance from each eve to the reference eve of that particuar
risk factor.A statistica tests were two-sided, and a P-vaues ess than
.05 were considered statisticay significant. Statistica anayses and
modeing were performed using Stata 10.0 (StataCorp, Coege
Station, TX) and SAS 9.2 (SAS Institute Inc, Cary, NC).
rsus
Risk of Hepatocellular Carcinoma in the Subcohort
Without HCV Test
Among 298 051 subjects in the subcohort without HCV test, 1252
incident hepatoceuar carcinoma occurred. We used data from
this subcohort to deveop four risk prediction modes. In a mod-
es, mae sex and oder age (4059 and 60 years) were statisticay
significanty associated with increased risk of hepatoceuar carci-
noma (Tabe 1). We presented risk factors besides sex and age in
the subsequent text.
Risk predictors that were statisticay significanty associated
with increased or decreased risks of hepatoceuar carcinoma in
mutivariabe modes are shown inTabe 1. In mode 1 (heath his-
tory ony), statisticay significanty increased risks were associatedwith smoking (19.9 vs 0 pack-years, HR = 1.35, 95% CI = 1.13 to
1.61; 10 vs 0 pack-years, HR = 1.32, 95% CI = 1.13 to 1.55), regu-
ar acoho drinking (reguar [consumed 2 drinks/day on 3 days/
week] vs none or occasiona [consumed
7/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
5/13
7/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
6/13
Vol. 104, Issue 20 | October 17, 20121604 Articles | JNCI
HR = 43.0, 95% CI = 27.8 to 66.60). In mode 3, a above vari-
abes were associated with a statisticay significanty increased risk,
except for ALT, and physica activity was not associated with a sta-
tisticay significanty decreased risk. In mode 4, HBV+ status was
associated with a statisticay significanty increased risk (HBV+ vs
HBV, HR = 4.04, 95% CI = 3.24 to 5.04). Again, increasing AFP
eve at or above 2.5 ng/mL was associated with statisticay sig-
nificanty increasing risk (2.54.9 vs
7/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
7/13
JNCI | Articles 1605jnci.oxfordjournals.org
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
A
Sensitivity
Model 2: AUC=0.900
Model 3: AUC=0.900
Model 4: AUC=0.918
Model 1: AUC=0.807
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
B
Sensitivity
Model 2: AUC=0.912
Model 3: AUC=0.913
Model 4: AUC=0.927
Model 1: AUC=0.793
Model 5: AUC=0.933
Figure 1. Discriminatory accuracy of the models. Discriminatory accuracy for predicting the development of hepatocellular carcinoma within10 years was assessed by constructing receiver operating characteristic (ROC) curves and calculating the area under the curve (AUC). A) Subcohortwithout hepatitis C virus (HCV) test. Four models were developed in this subcohort: model 1 was based on health history; model 2 was based ontransaminase only; model 3 was based on health history and transaminase; and model 4 was based on health history, transaminase level, alfa-fetoprotein level, and hepatitis B virus (HBV) status. B) Subcohort with HCV test. A fifth model was added to include the HCV status.
7/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
8/13
Vol. 104, Issue 20 | October 17, 20121606 Articles | JNCI
Figure 2. Internal calibration of the risk prediction models. Calibrationdetermined the extent of agreement between predicted and observedevents in 10 years, and then a cross-validated calibration plot was gen-erated for the different models. The dashed line indicates the referenceline for an ideal model. Solid circles mark the apparent predictions
for each decile, and the cross-validated predictions for each decile aremarked by a cross symbol. Vertical bars indicate 95% confidence inter-vals around the apparent prediction. Model 1 was based on healthhistory only; model 2 was based on transaminase only; model 3 wasbased on health history and transaminase; model 4 was based on
7/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
9/13
JNCI | Articles 1607jnci.oxfordjournals.org
of AST (2539 IU/L, score = 5; 4059 IU/L, score = 9; 60 IU/L,
score = 13) (Suppementary Tabe 1, avaiabe onine).
The probabiity of deveoping hepatoceuar carcinoma (ie,
absoute risk) in 5 and 10 years, as a function of increasing risk score,
in a modes in the two subcohorts is shown in Suppementary
Figure 2 (avaiabe onine).
Application of Risk Score and Prediction Power of
Transaminases
We appied the modes to predict probabiity of deveoping hepato-ceuar carcinoma in 5 years and 10 years using eight hypothetica
exampes (Tabe 4). These exampes are representative of individuas
from genera popuation with a range of risk profies. For exampe,
a 60-year-od mae with abnorma transaminases, AST (60 IU/L),
and ALT (30 IU/L), without considering any other risk factors, as
in Exampe 1, woud have a hepatoceuar carcinoma risk of 7.3%
(95% CI = 6.5% to 8.5%) and 15.5% (95% CI = 13.6% to 17.3%)
in 5 and 10 years, respectivey, according to mode 2. The same indi-
vidua in Exampe 2, when positive for HBV, woud have a 21.4%
(95% CI = 17.5% to 23.8%) and 38.2% (95% CI = 34.1% to 42.0%)
risk of cancer deveopment in 5 and 10 years, respectivey, according
to mode 4. When this individua is aso positive for HCV, as in
Exampe 3, his risk woud increase to 77.0% (95% CI = 69.1% to
82.8%) and 97.1% (95% CI = 94.7% to 98.4%) in 5 years and
10 years, respectivey, according to mode 5. This atter high risk
coud be substantiay attenuated to 44.6% (95% CI = 37.7% to
50.8%) and 75.8% (95% CI = 69.4% to 80.9%), respectivey, as in
Exampe 4, if ifestye risks were modified by smoking and drinking
cessation and by engaging in physica activity and improving dia-
betes management. In Exampes 5 and 6, we compared an individua
who was HBV+ with norma transaminase eves to someone whowas HBV with abnorma transaminase eves. The HBV individua
with abnorma transaminase eves had a much higher hepatoceu-
ar carcinoma risk compared with the individua with HBV posi-
tivity aone (5-year absoute risk = 5.1% [95% CI = 4.1% to 6.1%]
vs 0.1% [95% CI = 0.1% to 0.2%]; 10-year absoute risk = 11.8%
[95% CI = 9.9% to 13.6%] vs 0.3% [95% CI = 0.2% to 0.3%]).
Exampe 7 shows the substantia benefit of ifestye modification for
the high-risk individua in Exampe 6, which woud reduce his can-
cer risk by more than 50%. Exampe 8, an individua with HCV+
status, woud have a simiar risk to the individua in Exampe 5 with
HBV+, as ong as transaminases were norma.
health history, transaminase, alfa-fetoprotein level, and hepatitis Bvirus status; and in model 5, hepatitis C virus (HCV) status was added
to model 4. A) Subcohort without HCV test. B) Subcohort with HCVtest.
Figure 2. (Continued)
http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs372/-/DC1http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs372/-/DC1http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs372/-/DC1http://-/?-http://-/?-http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs372/-/DC1http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs372/-/DC1http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs372/-/DC17/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
10/13
Vol. 104, Issue 20 | October 17, 20121608 Articles | JNCI
Table4.
Applicationofthepredictionmodelstohypotheticalindividualswithdifferentriskprofiles*
Riskfactors
Example
1
Example2
Example
3
Example4
Example
5
Example6
Examp
le7
Example8
Transamin
ase
only
HBV+
HBV+andH
CV+
Lifestyleimproved,
HBV+,andHCV+
Normal
transamin
ase
butHBV
+
Abnormal
transaminase
butHBV
Lifestyleim
proved,
abnormaltransaminase
butHBV
Normal
transaminase
butHCV+
Age,y
60
60
60
60
60
60
60
60
Sex
Male
Male
Male
Male
Male
Male
Male
Male
AST,IU/L
60
60
60
60
20
60
60
20
ALT,
IU/L
30
30
30
30
20
30
30
20
Smoking,pack-years
10
10
0
10
10
0
10
Regulardrinking
Yes
Yes
No
Yes
Yes
No
Yes
Physicallyactive
No
No
Yes
No
No
Yes
No
Diabetes
Yes
Yes
No
Yes
Yes
No
Yes
AFP,ng/mL
7
10
10
2
10
10
2
HBVstatus
+
+
+
+
HCVstatus
+
+
+
Predictedprobabilityin
5y(95%CI),
%
7.3(6.5
to
8.5
)
21.4(17.5to23.8
)
77.0(69.1
to
82.8
)
44.6
(37.7to50.8
)
0.1
(0.1
to
0.2
)
5.1
(4.1
to6.1
)
2.1
(1.7t
o4.1
)
0.1
(0.1
to0.1
)
Predictedprobabilityin
10y(95%CI),
%
15.5
(13.6to17.3
)
38.2
(34.1
to42.0
)
97.1(94.7
to
98.4
)
75.8
(69.4
to80.9
)
0.3
(0.2
to
0.3
)
11.8
(9.9
to13.6
)
4.9
(2.5
to5.7
)
0.2
(0.2
to0.3
)
*
Themodelswereappliedtocalculatepro
babilityofdevelopinghepatocellularcarcinomaineighthypotheticalindividualswithdifferentriskprofiles(ie,
differentAST,ALTlevels,
lifestylefactorssuchassmoking,
drinking,physicalactivity,andhepatitisB
andCstatus)whorepresentedtheriskprofiles
inthegeneralpopulation.
Lifestyleimprovedm
eansquitsmoking,stopdrinking,orbecomingp
hysicallyactive.
Predicted
probabilitywascalculatedfromabsoluterisk.
ASTorALTlevelsgreaterthan25IU/Lwere
consideredabnormal.
AST=serumaspartatetransaminase;ALT=serumalaninetransaminase
;AFP=alfa-fetoprotein;
HBV=hepatitisBvirus;HCV=hepatitisCvirus;CI=confidenceinterval;=notapplicable.
ThisindividualonlyhadASTandALTmea
sured.
NodataonhealthhistoryvariablesorHB
VorHCVstatus.Thevaluesofthesevariablesw
ereleftblank.
7/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
11/13
JNCI | Articles 1609jnci.oxfordjournals.org
We aso anayzed the distribution of hepatoceuar carcinomaincidence by transaminase eves and by HBV and HCV status
(Tabe 5). Among 130 543 subjects in the subcohort with HCV
testing, 109 029 (83.5%) were negative for both HCV and HBV.
Of the 416 subjects who deveoped hepatoceuar cancer during an
average foow-up of 8.5 years, 66 (15.7%) were negative for both
HCV and HBV. Subjects with AST and ALT eves of 25 IU/L or
ower constituted 61.1% (79 762 of 130 543) of the HCV-tested
subcohort, and 7% (29 of 416) of subjects who experienced inci-
dent cancers had AST and ALT eves of 25 IU/L or ower. Incident
hepatoceuar cancers were detected in 4.6% (19 of 416) individu-
as with both HBV+ and HCV+ status. For those positive for HBV
or HCV, ony 37.2% (48 562 of 130 543) and 33.4% (43 601 of130 543) subjects, respectivey, were aware of their carrier status
(Tabe 5).
Dsusson
To date, pubished prediction toos are ony avaiabe for high-risk
chronic HBV carriers. In this study, we deveoped prediction mod-
es for hepatoceuar carcinoma based on data routiney coected
in a typica office visit with the goa to provide a simpe, efficient,
and widey avaiabe too to identify and quantify cancer risk in the
average-risk popuation. Our resuts showed that the mode with
transaminase aone was best abe to predict hepatoceuar carci-noma. Because this mode is abe to predict hepatoceuar carci-
noma risk with high prediction accuracy without knowing HBV
or HCV, it has great potentia to be transated into cinica use for
genera pubic.
HBV and HCV are we-known risk factors for hepatoceuar
carcinoma, but in this study, transaminase (AST or ALT) eve of 25
IU/L or higher were found be independent risk factors for hepato-
ceuar carcinoma with a inear doseresponse trend. The stepwise
prediction modes, invoving testing for AST or ALT initiay, are
simpe for cinicians to impement in their daiy practice. When
routiney coected AST exceeded 25 IU/L, the risk of hepatoce-
uar carcinoma increased exponentiay with increasing concentra-tions of AST. Because of the increased risks associated with AST or
ALT concentrations of 25 IU/L or higher, such a finding shoud
have triggered further testing for HBV, HCV, or AFP, to yied a
more compete picture. The mode using transaminases aone had
a high prediction power, with an AUC vaue of 0.912, which was
statisticay significanty better than those testing for HBV, HCV,
or AFP aone (see Suppementary Figure 1, avaiabe onine). On
the other hand, subjects with AST or ALT concentrations ess than
25 IU/L, considered norma in this study, can be spared from fur-
ther testing. Athough this norma group contributed 7% of a
hepatoceuar carcinoma incidences in this cohort, they occurred
Table 5. Distribution of study subjects, their awareness, and hepatocellular carcinoma incidence by AST or ALT values and by HBV andHCV status*
Transaminase level
Subjects, No. (%)
HCV and HCV and HCV+ and HCV+ and
Total HBV HBV+ HBV HBV+
All subjects in the HCV subcohort 130 543 (100) 109 029 (83.5) 18 155 (13.9) 3054 (2.3) 305 (0.2)
AST, IU/L
7/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
12/13
Vol. 104, Issue 20 | October 17, 20121610 Articles | JNCI
among neary two-thirds of the genera adut popuation (61%).
The risk eve in this group, with approximatey four incidences per
100 000 person-years, is sufficienty ow and may not be accorded
priority in further testing, because two-thirds of them, if pursued,
woud find their HBV or HCV status negative. At 25 IU/L or
greater, the fase-negative rate for AST was 12.7%. This is sma
compared with the fase-negative rate for HBV (51.7%) or HCV
(59.1%), as the mode shoud not miss those at high risk. Athough
the fase-positive rate for AST (24.2%) was higher than the cor-responding rate for HBV (14.3%) or HCV (2.5%), this difference
is not of much practica importance, because positive AST wi be
foowed-up with HBV or HCV test in this mode.
To our knowedge, this is the first mode that assesses the hepa-
toceuar carcinoma risk of apparenty heathy individuas visiting
their primary care physicians office for a heath checkup. In the it-
erature, prediction toos are ony avaiabe for high-risk individuas,
such as known chronic HBV carriers (79) or chronic HCV car-
riers (10,11). Modes for these high-risk individuas require more
cinica data and more sophisticated data to estimate risks; none
of them address the average- or unknown-risk genera popuation.
A mode for subjects at average or unknown risk is more vauabethan one for subjects at high risk, because detaied cinica data are
readiy avaiabe for the high-risk individuas but much ess avai-
abe for those at average or unknown risk. In this arge cohort, two-
thirds of those with positive HBV or HCV status, both high-risk
groups that avaiabe modes attempt to target, were not aware that
they were carriers (Tabe 5); thus, existing prediction modes were
of itte use for these groups.
The versatiity of the prediction mode we present is high-
ighted by the fact that it is usefu for both the average-risk genera
pubic and high-risk individuas. None of the currenty avaiabe
modes coud assess both HBV and HCV subjects, a arge group
contributing 3040% of iver cancer cases reported in western
popuations (6). In Taiwan, more than 1000 new hepatoceuarcancers coud be estimated to occur annuay in HBV and HCV
individuas (15). This group, constituting 83.5% of our cohort, is
commony overooked cinicay (3), and yet an individua in this
group is estimated to have a 10-year absoute risk of 11.8%, based
on our prediction mode (Exampe 6 inTabe 3). Existing modes,
moreover, were not abe to accuratey assess and may, therefore,
underestimate the risk of cancer in HBV+ and HCV+ patients, a
group contributing neary 5% of a cancer incidences in our study.
Subjects in this group coud have a 10-year risk as high as 97.1%,
as in Exampe 3 estimated by our mode. Thus, the simpe too we
present here has much wider appicabiity than other modes cur-
renty avaiabe. More importanty, this versatiity is accompishedwith high efficiency, by reying on common cinica data that are
often readiy avaiabe. That AST and ALT vaues aone showed
AUC exceeding 0.912 was remarkabe. These tests are inexpensive
and routiney coected in daiy medica practice but have not been
put into use for risk prediction in average-risk settings. Adoption
of this mode can make prediction of hepatoceuar carcinoma
a routine cinica activity among subjects in any category of risk.
Modes 4 and 5 showed the independent effects of AST or ALT on
iver cancer, even in the presence of HBV or HCV, but the effect
was appreciaby attenuated from mode 3. For exampe, the hazard
ratio for AST concentration of 4059 IU/L was 13.93 in mode 3
but reduced to 8.59 in mode 4 when HBV status was known. It was
further reduced to 6.31 in mode 5 when both HBV and HCV sta-
tus were considered (Tabe 2). In other words, HBV or HCV had
an impact on the predictive abiity of the transaminases, and there-
fore they were somewhat associated, but transaminase remained
a major statisticay significant predictor even in the presence of
hepatitis carrier status.
Athough this mode was mainy intended to predict cancer
incidence, we found that it was aso abe to predict mortaity fromiver-reated diseases (hepatoceuar carcinoma and iver cirrhosis)
with simiar precision (AUC = 0.93; data not shown). This find-
ing is an added benefit, and the identica resuts obtained (data not
shown) provided additiona confidence in the vaidity of the mode.
Our prediction mode is vauabe for risk identification and risk
communication. Subjects at high risk of cancer shoud be propery
informed in a timey manner of their reative and absoute risks.
However, the mode has vaue we beyond risk communication
because it is aso abe to estimate the potentia for risk mitigation
associated with heathy ifestye choices. Our mode identified four
statisticay significant risk factors from heath history, namey
smoking, drinking, physica inactivity, and diabetes, with a ofthese having bioogicay reevant and statisticay significant risks,
as previousy reported (13,22,23). Our data show that iver cancer
risk can potentiay be reduced by 3350% through modification
of one or more of these four risk factors. Counseing for ifestye
changes woud add important vaue to the risk prediction process.
These benefits can be seen not ony in high-risk individuas with
positive HBV or HCV status (Exampes 3 and 4 inTabe 3 show a
reduction from 77% to 45% in the 5-year risk) but aso in average-
risk individuas negative for HBV or HCV (Exampes 6 and 7 show
a reduction of simiar size, from 5.1% to 2.1%). Furthermore, in
addition to reducing risk for iver cancer, modifying and eiminat-
ing these ifestye behaviors can have a major impact on a-cause
mortaity (12,13), a fact often negected by cinicians. Reducinga-cause mortaity is ceary as important as reducing iver cancer.
Thus, a further vauabe aspect of this simpe prediction mode is
that it may hep to educate and motivate an individua at high risk
to pursue a wider range of various risk reduction options. These
options range from treating carriers with interferon-ike therapy
(6) to reducing cancer risk by eiminating associated risk factors to
reducing a-cause mortaity.
The high predictive power of eevated transaminase eves
of AST or ALT at 25 IU/L or higher for hepatoceuar cancer
risk, athough not a nove finding, has nevertheess not been fuy
appreciated in the cinic. This finding arose by examining the can-
cer risk at every 5-unit interva across the entire AST and ALTdistribution. The cutoff eve of 25 IU/L is worth remembering,
because it is substantiay ower than the 40 IU/L commony cited
by aboratory standards as the ULN. Neary one-fifth (17.8%) of
the adut popuation in our cohort had an AST between 25 and 39
IU/L, a eve commony dismissed as high norma, and yet their
iver cancer risk was increased by as much as 3.6-fod. Once AST
exceeded 40 IU/L, the risk score given was as arge as, or equiva-
ent to, that of the HBV+ or HCV+ individuas in mode 5. We
aso found that the predictive power of AST and ALT aone for
iver cancer was higher than that of isoated HBV, HCV, or AFP
(Suppementary Figure 1, avaiabe onine). Finay, athough the
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs372/-/DC1http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs372/-/DC1http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-7/31/2019 JNCI J Natl Cancer Inst 2012 Wen 1599 611
13/13
JNCI | Articles 1611jnci.oxfordjournals.org
abnorma ALT has been commony recognized to pay an impor-
tant roe, our data showed that the risk conferred by abnorma
AST was higher than ALT (6.3- to 8.3-fod increase vs 1.9-fod
increase;Tabe 2, mode 5).
There are severa imitations of this study. First, athough our
modes demonstrated an exceent eve of goodness of fit and dis-
criminatory abiity, additiona vaidation in externa popuations
is advised. To some extent, the simiar resuts in the two different
subcohorts, each of which had a arge sampe size, provide inter-na vaidity to our findings, but externa confirmation is needed.
Second, our cohort is drawn from participants engaged in a medi-
ca screening program beonging to an above-average socioeco-
nomic status. This may imit the generaizabiity of our findings.
Nevertheess, the hazard ratios deveoped in this study were inter-
nay standardized and free from seection bias. The sampe sizes of
the two subcohorts were sufficienty arge, representing 3% of the
tota adut popuation in Taiwan. Most resuts derived from these
two subcohorts, such as C-index or reative risks, were simiar.
Third, ony AST or ALT data from the initia visit were used for the
mode, and tempora changes were not considered. Nevertheess,
the predictive power of a singe transaminase test was reinforced inthis study, as in other studies invoving this cohort (12,13).
In summary, the use of transaminase data was best abe to pre-
dict hepatoceuar carcinoma risks, with AUC vaue of 0.90 or
higher. Athough HBV and HCV are we-known risk factors, AST
or ALT concentrations of 25 IU/L or higher had independent and
higher predictive power, even among unknown or HBV or HCV
subjects, and shoud trigger further testing. This simpe too for
the genera pubic more accuratey assesses risk even among groups
previousy thought to be at ow or average risk and may be hepfu
to educate and motivate individuas to pursue options beneficia in
reducing their risk of iver cancer and a-cause mortaity.
rfns1. Kuehn BM. Report: too itte surveiance, treatment for US patients with
hepatitis B and C.Jama. 2010;303(8):713714.
2. Hepatitis and Liver Cancer: A Nationa Strategy for Prevention and
Contro of Hepatitis B and C.www.iom.edu/virahepatitis. 2010.
3. Bonski W, Kotyar DS, Forde KA. Non-vira causes of hepatoceuar car-
cinoma. World Journal of Gastroenterology. 2010;16(29):36033615.
4. Chen JG, Zhang SW. Liver cancer epidemic in China: past, present and
future. Semin Cancer Biol. 2011;21(1):5969.
5. U.S. Preventive Services Task Force. Screening for Hepatitis C in Adults.
http://www.uspreventiveservicestaskforce.org/uspstf/uspshepc.htm .
Accessed Juy 26, 2012.
6. E-Serag HB. Current concepts hepatoceuar carcinoma.N Engl J Med.
2011;365(12):11181127.
7. Yang HI, Sherman M, Su J, et a. Nomograms for risk of hepatoceuar
carcinoma in patients with chronic hepatitis B virus infection.J Clin Oncol.
2010;28(14):24372444.
8. Wong VWS, Chan SL, Mo F, et a. Cinica scoring system to predict
hepatoceuar carcinoma in chronic hepatitis B carriers. J Clin Oncol.
2010;28(10):16601665.
9. Yuen MF, Tanaka Y, Fong DYT, et a. Independent risk factors and pre-
dictive score for the deveopment of hepatoceuar carcinoma in chronic
hepatitis B. J Hepatol. 2009;50(1):8088.
10. Ikeda K, Arase Y, Saitoh S, et a. Prediction mode of hepatocarcinogen-
esis for patients with hepatitis C virus-reated cirrhosis. Vaidation with
interna and externa cohorts.J Hepatol. 2006;44(6):10891097.
11. Masuzaki R, Tateishi R, Yoshida H, et a. Prospective risk assessment for
hepatoceuar carcinoma deveopment in patients with chronic hepatitis C
by transient eastography.Hepatology. 2009;49(6):19541961.
12. Wen CP, Cheng TYD, Tsai MK, et a. A-cause mortaity attributabe to
chronic kidney disease: a prospective cohort study based on 462 293 aduts
in Taiwan.Lancet. 2008;371(9631):21732182.
13. Wen CP, Wai JPM, Tsai MK, et a . Minimum amount of physica activity
for reduced mortaity and extended ife expectancy: a prospective cohort
study.Lancet. 2011;378(9798):12441253.
14. Chang MH, You SL, Chen CJ, et a. Decreased incidence of hepatoceuar
carcinoma in hepatitis B vaccines: a 20-year foow-up study.J Natl Cancer
Inst. 2009;101(19):13481355.
15. Department of Heath. Statistics of causes of death. http://www.doh.gov.tw/
EN2006/DM/DM2.aspx?now_fod_ist_no=9256&cass_no=390&eve_
no=2. Accessed Juy 26, 2012.
16. Grambsch PM, Therneau TM. Proportiona hazards tests in diagnostics
based on weighted residuas. Biometrika. 1994;81:551526.
17. Prati D, Taioi E, Zanea A, et a. Updated definitions of heathy ranges for
serum aanine aminotransferase eves.Ann Intern Med. 2002;137(1):110.
18. Heagerty PJ, Lumey T, Pepe MS. Time-dependent ROC curves
for censored surviva data and a diagnostic marker. Biometrics.
2000;56(2):337344.
19. DAgostino RD, Nam BH. Evauation of performance of surviva ana-
ysis modes: discrimination and caibration measures. In: Baakrishnan N,
Rao CR, eds.Handbook of Statistics. Vo 23. Amsterdam, the Netherands:
Esevier; 2004:125.
20. Hosmer DW, Lemeshow S.Applied Survival Analysis: Regression Modeling of
Time to Event Data. New York, NY: John Wiey & Sons, Inc; 1999.
21. Suivan LM, Massaro JM, D'Agostmo RB. Presentation of mutivariate
data for cinica use: the Framingham study risk score functions. Stat Med.
2004;23(10):16311660.
22. Kuper H, Tzonou A, Kakamani E, et a. Tobacco smoking, acoho con-
sumption and their interaction in the causation of hepatoceuar carcin-
oma. Int J Cancer. 2000;85(4):498502.
23. E-Serag HB, Hampe H, Javadi F. The association between diabetes and
hepatoceuar carcinoma: a systematic review of epidemioogic evidence.
Clin Gastroenterol Hepatol. 2006;4(3):369380.
Fundng
This study was supported in part by Taiwan Department of Heath Cinica Tria
and Research Center of Exceence (grant number DOH101-TD-B-111-004 to
CPW), The University of Texas MD Anderson Cancer Center Research Trust
(to XW), and Center for Transationa and Pubic Heath Genomics, Duncan
Famiy Institute for Cancer Prevention and Risk Assessment, The University of
Texas MD Anderson Cancer Center (to XW).
Nos
The funders had no roe, and the authors are responsibe for the study design,
data coection, data anaysis, interpretation, writing of the manuscript, and the
decision to submit the manuscript for pubication.
Affiliations of authors: Institute of Population Science, National Health
Research Institutes,Zhunan, Taiwan (C-PW, YCY, MKT); China Medical
University Hospital, Taichung, Taiwan (C-PW, YCY, MKT, CYH); Departments
of Epidemiology (JL, CE, MH, YY, XW), Gastroenterology, Hepatology and
Nutrition (LM), and Division of Cancer Prevention and Population Sciences
(EH), The University of Texas MD Anderson Cancer Center, Houston, TX; MJ
Health Management Institution, Taipei, Taiwan (CKT).
http://-/?-http://-/?-http://-/?-http://www.iom.edu/viralhepatitishttp://www.uspreventiveservicestaskforce.org/uspstf/uspshepc.htmhttp://www.doh.gov.tw/EN2006/DM/DM2.aspx?now_fod_list_no=9256%26class_no=390%26level_no=2http://www.doh.gov.tw/EN2006/DM/DM2.aspx?now_fod_list_no=9256%26class_no=390%26level_no=2http://www.doh.gov.tw/EN2006/DM/DM2.aspx?now_fod_list_no=9256%26class_no=390%26level_no=2http://www.doh.gov.tw/EN2006/DM/DM2.aspx?now_fod_list_no=9256%26class_no=390%26level_no=2http://www.doh.gov.tw/EN2006/DM/DM2.aspx?now_fod_list_no=9256%26class_no=390%26level_no=2http://www.doh.gov.tw/EN2006/DM/DM2.aspx?now_fod_list_no=9256%26class_no=390%26level_no=2http://www.uspreventiveservicestaskforce.org/uspstf/uspshepc.htmhttp://www.iom.edu/viralhepatitishttp://-/?-http://-/?-http://-/?-