Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
Transplantation DIRECT 2022 www.transplantationdirect.com 1
Magnetic Resonance Imaging Predictors of
Hepatocellular Carcinoma Progression and
Dropout in Patients in Liver Transplantation
Waiting List
Azarakhsh Baghdadi, MD,1,* Harry T. Luu, MD, MPH,2,* Mohammadreza Shaghaghi, MD,1
Maryam Ghadimi, MD,1 Cem Simsek, MD,3 Ziyi Xu, BA,2 Bita Hazhirkarzar, MD,1 Mina Motaghi, MD,1
Muhammad Hammami, MD,4 Jeanne M. Clark, MD,2,5,6 Ahmet Gurakar, MD,2 Ihab R. Kamel, MD, PhD,1
and Amy K. Kim, MD2
ISSN: 2373-8731
DOI: 10.1097/TXD.0000000000001365
Received 15 March 2022. Revision received 11 July 2022.
Accepted 13 July 2022.
1 Department of Radiology and Radiological Sciences, Johns Hopkins University
School of Medicine, Baltimore, MD.
2 Department of Medicine, Gastroenterology and Hepatology, Johns Hopkins
University School of Medicine, Baltimore, MD.
3 Brigham and Women’s Hospital, Division of Gastroenterology, Boston, MA.
4 University of Maryland Medical Center, Division of Gastroenterology and
Hepatology, Baltimore, MD.
5 Department of Medicine, General Internal Medicine, Johns Hopkins University
School of Medicine, Baltimore, MD.
6 Department of Epidemiology, Johns Hopkins School of Public Health,
Baltimore, MD.
This work was supported by NCI K08-CA237624.
This study is approved by our local institutional review board NA_00068847.
A.K.K. is a consultant to Astrazeneca and Exelixis. The other authors declare no
conflicts of interest.
A.B., H.L., and I.R.K. contributed to conception and design, acquisition, writing
of the article. M.S. performed acquisition, analysis of data, writing of the article.
M.G. and B.H. performed acquisition of data. C.S. contributed to design and
Liver Transplantation
Background. With the rising incidence of hepatocellular carcinoma (HCC), more patients are now eligible for
liver transplantation. Consequently, HCC progression and dropout from the waiting list are also anticipated to rise.
We developed a predictive model based on radiographic features and alpha-fetoprotein to identify high-risk patients.
Methods. This is a case-cohort retrospective study of 76 patients with HCC who were listed for liver transplanta-
tion with subsequent liver transplantation or delisting due to HCC progression. We analyzed imaging-based predictive
variables including tumor margin (well- versus ill-defined), capsule bulging lesions, volumetric analysis and distance to
portal vein, tumor numbers, and tumor diameter. Volumetric analysis of the index lesions was used to quantify index
tumor total volume and volumetric enhancement, whereas logistic regression and receiver operating characteristic curve
(ROC) analyses were used to predict the main outcome of disease progression. Results. In univariate analyses, the
following baseline variables were significantly associated with disease progression: size and number of lesions, sum
of lesion diameters, lesions bulging the capsule, and total and venous-enhancing (viable) tumor volumes. Based on
multivariable analyses, a risk model including lesion numbers and diameter, capsule bulging, tumor margin (infiltrative
versus well-defined), and alpha-fetoprotein was developed to predict HCC progression and dropout. The model has an
area under the ROC of 82%, which was significantly higher than Milan criteria that has an area under the ROC of 67%.
Conclusions. Our model has a high predictive test for patient dropout due to HCC progression. This model can
identify high-risk patients who may benefit from more aggressive HCC treatment early after diagnosis to prevent dropout
due to such disease progression.
(Transplantation Direct 2022;8: e1365; doi: 10.1097/TXD.0000000000001365).
acquisition. Z.X. contributed to acquisition of data and analysis. M.M. and B.H.
performed acquisition of data. J.M.C. performed analysis, interpretation of data,
and critical revision. A.K.K. contributed to conception and design, acquisition,
analysis, and writing of the article.
Supplemental digital content (SDC) is available for this article. Direct URL
citations appear in the printed text, and links to the digital files are provided in the
HTML text of this article on the journal’s Web site (www.transplantationdirect.
com).
Correspondence: Amy K. Kim, MD, Department of Medicine, Division of
Gastroenterology and Hepatology, Johns Hopkins University School of Medicine,
720 Rutland Ave. Ross 918, Baltimore, MD 21205. (akim97@jhmi.edu).
Copyright © 2022 The Author(s). Transplantation Direct. Published by Wolters
Kluwer Health, Inc. This is an open-access article distributed under the terms of
the Creative Commons Attribution-Non Commercial-No Derivatives License 4.0
(CCBY-NC-ND), where it is permissible to download and share the work provided
it is properly cited. The work cannot be changed in any way or used commercially
without permission from the journal.
2 Transplantation DIRECT ■ 2022 www.transplantationdirect.com
Hepatocellular carcinoma (HCC) has become the sec-
ond fastest-growing cancer in the United States,1 with
600 000 new patients diagnosed worldwide (20 000 in the
United States).2 With the rising incidence of HCC, more patients
with HCC are being listed for liver transplantation, cementing
it as one of the most common listing diagnoses.3 However, not
all patients with HCC listed for liver transplants can withstand
the waitlist duration, which can vary up to 2 y depending on
region and blood type. Considering the current waiting period
with exception criteria, 20%–35% of US patients eventually
drop from the list because of disease progression.4
Dropout from the waiting list due to HCC progression
occurs despite ongoing bridging therapy with liver-directed
treatments such as transarterial chemo-embolization (TACE)
and radio-embolization.5 This suggests there is a substantial
heterogeneity of the HCC population on the transplant list,
despite strict criteria such as the Milan criteria that limit the
tumor burden to a specic size and number.6 Eligibility is also
complicated by a downstaging process that includes baseline
tumors beyond Milan criteria but with reduction in tumor
burden due to good treatment response.7
Understanding that HCC is a heterogeneous disease with
various risk factors of disease progression,8 there is an urgent
need to identify those with more aggressive features at risk
of progression and dropout earlier in the evaluation process.
This is imperative as more than 90% of HCC emerges in cir-
rhosis and that progressive decline in liver function in these
patients often restricts HCC treatment options.9 Given that
systemic therapies in HCC have advanced signicantly, includ-
ing immunotherapy and other tyrosine kinase inhibitors,10-13
early identication of high-risk patients with HCC using an
accurate risk score may now allow for concurrent systemic
therapy with standard locoregional treatments before the fur-
ther decline of their liver function.
Previous studies have associated certain risk factors with a
higher dropout rate, including liver decompensation, tumor size,
and the number of lesions.14,15 However, large national database
studies are limited to reported 2D tumor dimensions based on
standard clinical practice.16 Recent advances in high-resolution
imaging now allow 3D reconstruction of CT and MRI for
volumetric analysis and functional imaging measures such as
diffusion-weighted imaging and viable tumor volume.17-19 These
clinical tools are yet to be incorporated routinely into prognos-
tic models. Our group has previously shown the predictive use
of volumetric imaging analysis in HCC in predicting overall sur-
vival.20 Therefore, this advanced postprocessing of radiographic
data from baseline imaging may help identify patients with a
high dropout risk due to HCC progression.
We set out to test whether more detailed radiographic
parameters, in addition to standard HCC measurements,
could better predict which patients with HCC progression
while on the liver transplant list would result in dropout from
the list. We hypothesized that such a predictive model would
include tumor border, volumetric analysis, proximity to a ves-
sel, and touching the capsule. A predictive model will identify
high-risk patients for broader or earlier aggressive treatments.
To our knowledge, this is the rst predictive model for HCC
progression while on the liver transplant waitlist.
MATERIALS AND METHODS
This single-center case-control study was HIPAA-
compliant and approved by our local institutional review
board. We identied 389 patients from the UNOS data-
base listed in our center for liver transplantation between
2010 and 2017, with the primary diagnosis of hepatobil-
iary malignancy. Of those, we reviewed 69 patients with a
conrmed diagnosis of HCC and transplant listing, with
subsequent removal from listing due to HCC progression.
Patients who were removed from the waitlist due to non–
cancer-related medical conditions, psychosocial reasons, or
lost follow-up were excluded. Among the included patients
were 34 cases with baseline MRI available in our system.
We randomly selected for comparison an additional 42
HCC patients who received liver transplantation during
the same period with available baseline MRI results. A total
of 76 patients met the selection criteria (Figure1). In this
study cohort, 69 out of 76 (91%) patients received TACE.
Among those 7 patients who did not receive locoregional
treatment, 6 were from liver transplantation group and 1
from progression group. Four patients received radiofre-
quency ablation, 2 from each group. Two patients from the
progression group had sorafenib toward the end of their
delisting date. No patients received immunotherapy.
HCC treatment history, transplant listing date, wait-
list dropout, and transplantation dates were retrieved from
electronic medical records. We obtained clinical data includ-
ing etiology of liver disease, alpha-fetoprotein (AFP) level,
and baseline labs at the time of referral to assess the sever-
ity of liver disease using model for end-stage liver disease
(MELD)with sodium (MELD-Na). Detailed imaging analysis
was performed as described below. Dropout from the waiting
list due to HCC progression was the primary outcome for
patients in our study.
Imaging Analysis
MRI was independently reviewed and analyzed by 1 senior
radiologist and 3 radiology research fellows, blinded to the
patient’s information. The largest tumor on baseline imag-
ing was selected as the index lesion for all patients; the index
lesions were segmented on baseline imaging. Image analysis
included the total number of lesions, the diameter of the larg-
est lesion, the sum of diameters of all lesions, lesion bulging
outside the liver capsule, lesion touching the liver capsule,
lesion touching the portal vein, tumor margin (ill-dened
versus well-dened), tumor volume, and volumetric venous
enhancement.
Standalone prototype software (Parametric Toolbox, ver-
sion1; Siemens Healthcare, Malvern, PA) was used to obtain
portal venous enhancement maps from precontrast (P) and
portal venous phase (V) images. To calculate venous enhance-
ment maps,
V−P
P×100
, precontrast images were coregistered
to portal venous phase images by nonrigid 3D registration.
They were exported in Digital Imaging and Communications in
Medicine format. Subsequently, portal venous phase enhance-
ment maps were uploaded to MR OncoTreat prototype soft-
ware (MR OncoTreat, Siemens Healthcare, Princeton, NJ) for
volumetric analysis. The volume of the tumor and voxel-wise
histograms of enhancement values were calculated using the
software mentioned above.
Viable tumor volume was derived from venous enhance-
ment maps based on a threshold of viability for venous
enhancement. This method was described in a previ-
ous study.21 Voxels with enhancement values equal to
or lower than this threshold were considered necrotic,
and those above the threshold were deemed viable. The
© 2022 The Author(s). Published by Wolters Kluwer Health, Inc. 3Baghdadi et al
following equation was used to calculate viable volume:
Number of voxels showing enhancement above threshold
Total number of voxels ×Tumor volume
.
Statistical Analysis
Patients were grouped into those who underwent trans-
plantation and those who progressed. Categorical parameters
are presented as the number and percentage, and continuous
data as the median and interquartile range. The relation-
ship between variables of interest and group (transplant or
progression) was evaluated using the Student’s t-test and
Wilcoxon rank sum test when appropriate for continuous
variables and using the chi-squared test or Fisher’s exact
test for categorical variables. Logistic regression was used
to identify candidates for predictors of disease progression.
Multivariable logistic regression predicting disease progres-
sion was performed using the selected variables and AFP.
Collinearity and dependency among variables were examined,
and variables showing a high level of correlation and depend-
ency were removed from the nal model. For risk-scoring
analysis, continuous variables were dichotomized before per-
forming multivariable logistic regression. Cut-off values were
determined based on univariate analyses. From the multivari-
able model, the predicted risk was calculated using the fol-
lowing formula ¯
p=e(
b0
+
b1X1
+
b2X2
+...+
bnXn
)
1+e
(b0+b1X1+b2X2+...+bnXn)
,
in which b0 is
the constant and b1 through bn are the regression coefcients
(log of odds ratios), X1 through Xn are dichotomized predic-
tors of disease progression (Yes = 1 and No = 0), and the
numerator is the risk score. The model’s predictive perfor-
mance was evaluated by the Hosmer-Lemeshow goodness-
of-t test and the receiver operating characteristic (ROC)
analysis. ROC curves for all models were constructed22 and
compared statistically.15 Prediction model validation was
performed using 10-fold cross-validation, in which our
data were randomly divided into 10 subgroups. The model
was then tted with the data in the 9 subgroups, and the
remaining subgroup was used for validation. The analysis
was repeated 10 times, with each subgroup being used once
as the validation set. After each analysis, the AUC and the
root mean square error were calculated, and from these 10
statistics, the mean, SD, and 95% condence interval were
determined.23 All analyses were performed using StataCorp
2017 (College Station, TX: Statacorp LLC). P < 0.1 was con-
sidered suggestive signicance for inclusion in model building.
For all other analysis, the threshold for statistical signicance
was at P < 0.05.
RESULTS
Meeting the eligibility criteria were 76 patients, of whom
42 received liver transplants (transplant group), whereas 34
FIGURE 1. Patient selection criteria. Patients were first identified from UNOS database with the diagnosis of “hepatobiliary malignancy” and then
subsequently included based on MRI availability and if their delisting was due to HCC progression outside Milan criteria. HCC, hepatocellular carcinoma.
4 Transplantation DIRECT ■ 2022 www.transplantationdirect.com
experienced disease progression and were dropped from the
waitlist (progression group). Age, sex, MELD, and etiology of
liver disease were similar between the 2 groups. As expected,
more patients in the transplant group met the Milan criteria
at baseline imaging than those in the progression group (66%
versus 34%, P = 0.0004) (Table1). In the transplant group,
AFP trended to be lower than in the progression group, and
the difference was statistically signicant (P = 0.0008). For
those who received a liver transplant (transplant group), time
on the waitlist calculated from the day of listing to the day of
transplant was longer (269 d) than the duration from listing
to dropout (196 d) for those who dropped off the list before
a transplantable organ became available (progression group).
Duration from rst baseline imaging to transplant was 316 d,
whereas duration from baseline imaging to waitlist dropout
was 253 d. Patients who remained in the transplant group
were less likely (mean number 1.5 versus 2.5) to get locore-
gional treatment with TACE, despite remaining on the waitlist
longer than patients in the progression group who dropped
out sooner.
To identify the key radiographic differences between
the 2 groups, we used logistic regression analysis for each
variable. At the signicance level of P < 0.10, the follow-
ing baseline radiographic variables were associated with
HCC progression and waitlist dropout (Table2): number
of lesions, diameter of dominant lesion, sum of diameters
of all lesions, lesion bulging outside of liver capsule, tumor
margin, tumor volume and venous enhancement volume of
the dominant lesion, and Milan criteria. A tumor touching
the portal vein and MELD-Na were not associated with
progression. Based on this univariable analysis and the
exclusion of collinear and dependent variables, we selected
the number of lesions, the diameter of the dominant lesion,
capsule bulging lesion, tumor border, and AFP to create
a predictor model of HCC progression in the transplant
waitlist period.
Multivariable Analysis and Risk Prediction
To build a clinically useful predictor model and risk scor-
ing analysis, we rst created cut-off values for the continu-
ous variables. Cut-off values were determined based on our
univariate analysis: number of lesions >1, diameter of domi-
nant lesion >3.0 cm, and AFP >60 ng/dL. We performed a
multivariable logistic regression analysis (Table3). At the sig-
nicance level of P < 0.05, the association between disease
progression and the number of lesions and dominant lesion
diameter was still observed. The associations of tumor bulg-
ing, tumor margin, and AFP were attenuated. The direction
of all associations remained similar to the univariate analy-
sis, and the Hosmer-Lemeshow goodness-of-t test (P = 0.37)
showed that the prediction model was a good t. As Milan
criteria are the standard criteria for transplant eligibility at
the time of listing, we compared our prediction model with
TABLE 1.
Baseline clinical characteristics of patients who received liver transplantation vs those with disease progression
Transplanted (n = 42) Progressed (n = 34) P
Age, y, median (range) 63 (44–69) 63 (42–71) 0.757
Sex, n (%) Female 7 (43.8) 9 (56.2) 0.447
Male 35 (58.8) 25 (42.2)
Etiology, n (%) HCV 23 (51.1) 22 (48.9) 0.803
HBV 5 (100.0) 0 (0.0)
EtOH 4 (66.7) 2 (33.3)
HCV, EtOH 1 (20.0) 4 (80.0)
NAFLD 7 (58.3) 5 (41.7)
EtOH, NAFLD 1 (100.0) 0 (0.0)
Other 1 (50.0) 1 (50.0)
Milan criteria, n (%) Yes 39 (66.1) 20 (33.9) <0.001
No 3 (17.6) 14 (82.4)
MELD-Na, median (IQR) 8 (7–9) 9 (8–11) 0.075
AFP, ng/mL, median (IQR) 7.4 (4.7–16.3) 23.1 (9.8–122.2) <0.001
Waitlist duration, d, median (IQR) 316 (250–386) 253 (193–405) 0.212
Listing to end point, d, median (IQR) 269 (205–364) 196 (107–417) 0.210
Number of lesions, median (IQR) 1 (1–1) 2 (1–3) <0.001
Lesion diameter, cm, median (IQR) 2.5 (1.8–3.3) 3.7 (2.6–4.6) <0.001
Number of TACE, median (IQR) 1.5 (1–2) 2.5 (1–4) 0.004
AFP, alpha-fetoprotein; EtOH, alcoholic hepatitis; HBV, hepatitis B virus; HCV, hepatitis C virus; IQR, interquartile range; MELD-Na, model for end-stage liver disease with sodium; NAFLD, nonalcoholic
fatty liver disease; TACE, transarterial chemoembolization treatments.
TABLE 2.
Univariate analysis of clinical and radiographic variables
as predictors of disease progression and liver
transplantation waitlist dropout
Odds ratio 95% CI P
Milan criteria, yes vs no 0.11 0.028-0.427 0.001
MELD-Na, per point 1.07 0.960-1.202 0.210
AFP, per 1 ng/mL 1.01 0.999-1.020 0.078
Number of lesions, multiple vs solitary 5.63 1.960-16.147 0.001
Diameter of dominant lesion, (per 1 cm) 2.42 1.491-3.943 <0.001
Sum of diameters of all lesions, (per 1 cm) 2.20 1.487-3.243 <0.001
Capsule bulging, yes vs no 4.09 1.565-10.687 0.004
Touched portal vein, yes vs no 1.53 0.518-4.520 0.442
Tumor margin, ill-defined vs well-defined 2.57 0.940-7.010 0.066
Tumor volume, (per 1 mL) 1.04 1.008-1.074 0.014
Venous enhancement volume, (per 1 mL) 1.04 1.006-1.075 0.019
AFP, alpha-fetoprotein; CI, confidence interval; MELD-Na, model for end-stage liver disease with
sodium.
© 2022 The Author(s). Published by Wolters Kluwer Health, Inc. 5Baghdadi et al
Milan criteria for predicting HCC progression and dropout.
This current model improved the prediction of HCC progres-
sion by 15% compared with Milan criteria, which has area
under the ROC (AUROC) 0.67 (95% condence interval [CI]:
0.577-0.763). At 83% specicity, the sensitivity of the model
is 68% (Figure2). The model has a high discriminating power
for predicting disease progression (AUROC = 0.82, 95% CI:
0.728-0.919). Using a 10-fold cross-validation technique, the
model had a mean AUROC of 0.82 (95% CI: 0.633-0.866)
and a root mean squared error of 0.43, conrming a strong
predictive performance and relatively high accuracy (Figure
S1, SDC, http://links.lww.com/TXD/A448).
For clinical application, we then formulated an equation
for risk scoring based on the multivariable analysis and,
nally, the predicted risk analysis. The predicted risk for HCC
progression and waitlist dropout ranged from 83% to 95%, if
a patient had any of the 4 predictors (Table4). Based on this
risk score, a patient with 2 lesions, with the largest lesion hav-
ing a diameter of 2.8 cm, inltrative tumor margin, bulging
the capsule, and AFP of 90 µg/dL would have a risk score of
2.1 and predicted risk of progression of 89%.
DISCUSSION
Despite strict eligibility criteria and bridging therapy for
liver transplantation, patients with HCC remain at risk of
HCC progression and dropout from the waiting list. This
study demonstrated the use of additional radiographic data
combined with AFP to develop a predictive risk model for
HCC progression and dropout. We also show that 3D volu-
metric image data can be analyzed from standard high-reso-
lution MR imaging.
Liver transplant eligibility criteria have remained stringent
to improve patient outcomes, mainly focused on survival and
HCC recurrence after transplantation; limited studies have
assessed risk factors for patients dropping off the list due to
disease progression. We performed a comprehensive image
analysis to consider new radiographic markers associated
with HCC progression, including information about tumor
characteristics, which is more detailed than that reported in
national databases such as UNOS. We also evaluated base-
line imaging from all patients referred to liver transplantation,
including those outside Milan criteria, to reduce any bias of
previous treatment response. Before adopting Milan criteria,
Yao et al initially suggested a similar cut-off of tumor lesion
size and numbers as predictors for dropout.24 Since then, stud-
ies have associated AFP,25 multifocal HCC lesions, and MELD
as risk factors for overall dropout.26 Our ndings show simi-
lar MELDs between transplant and progression groups. This
may be due to our selection of dropout cases from HCC pro-
gression only, rather than other causes of liver decompensa-
tion or transplant-limiting medical conditions.
More accurate prediction of HCC progression and early
identication of these high-risk patients on the waitlist have
several important implications. First, the treatment landscape
of HCC has evolved with more effective systemic treatment
options, including new tyrosine-kinase inhibitors and immu-
notherapy options.27 Clinical trials are underway to investigate
TABLE 3.
Multivariable logistic regression analysis predicting
disease progression and liver transplantation waitlist
dropout (n = 76)
Odds ratio 95% CI P
Number of lesions, multiple vs solitary 5.88 1.715-20.180 0.005
Lesion diameter of dominant lesion, >3 cm
vs ≤3 cm
3.48 1.095-11.086 0.035
Ill-defined margin, yes vs no 1.36 0.385-4.810 0.632
Capsule bulging, yes vs no 2.29 0.731-7.152 0.155
AFP, >60 ng/mL vs ≤60 ng/mL 3.86 0.860-17.346 0.078
AFP, alpha-fetoprotein; CI, confidence interval.
FIGURE 2. Model performance in distinguishing patients who drop out from the waitlist due to HCC progression. Model (solid line) is compared
with Milan criteria (dotted line). HCC, hepatocellular carcinoma.
6 Transplantation DIRECT ■ 2022 www.transplantationdirect.com
combination treatments that include locoregional therapies
with systemic treatments for intermediate stage HCC and
are expected to impact our transplant patient population.28
At this critical junction, an accurate risk score for prognosis
can identify patients with HCC who may not benet from
locoregional treatment alone. These patients can potentially
start concurrent therapy with closer surveillance before the
further decline of their liver function.
Some key ndings merit further discussion. First, there
were no signicant differences in MELDs between the groups,
as noted in another study, suggesting that patients in these 2
groups had a comparable liver function at baseline. Second,
capsule bulging and tumor margin were associated with list
dropout. It has been well-known that exophytic lesions and
liver capsules are supplied by extrahepatic collateral ves-
sels,29,30 which may mitigate the embolization effect of TACE.
Capsular interruption and irregular tumor margins could
indicate microvascular invasion, a prognostic factor for post-
transplant recurrence and metastases.31
In contrast, volumetric analysis, including tumor vol-
ume and venous enhancement volume, was associated with
progression in univariate analysis but not in multivariable
analysis. This nding contradicts our previous studies that
associated volumetric venous enhancement with HCC histol-
ogy and patients’ overall survival in both HCC and intrahe-
patic cholangiocarcinoma.21,32 This may be due to the strong
effects of the number of lesions and lesion diameter and the
dichotomization of continuous variables in the current study
cohort. The discrepant nding in this study may include a
contribution by the overall smaller tumor volume in general
for patients awaiting liver transplantation compared with
more advanced stage HCC, or due to smaller sample size, and
importantly, limited follow-up duration (ie, transplant or pro-
gression exceeding Milan) compared with earlier studies.
Our study has some limitations. This is a single-center ret-
rospective study specic to the demographics in our region.
Sample size is relatively small, as patients in the progression
group were selected after conrmation that they had dropped
off the list because of HCC rather than other causes of drop-
out including social issues, lack of follow-up, or other medical
comorbidities. Furthermore, baseline tumor characteristics at
the time of referral may be variable per geographic region,
TABLE 4.
Risk score and predicted risk of disease progression and liver transplantation waitlist drop out using selected predictors
Number of lesions >1 Dominant lesion diameter >3 cm Ill-defined margin Capsule bulging AFP >60 ng/mL Predicted risk
5 out of 5 risk factors
Yes Yes Yes Yes Yes 97%
4 out of 5 risk factors
Yes Yes No Yes Yes 95%
Yes Yes Yes No Ye s 93%
Yes No Yes Yes Yes 89%
Yes Yes Yes Yes No 88%
No Yes Yes Yes Yes 83%
3 out of 5 risk factors
Yes Yes No No Yes 90%
Yes No No Yes Yes 86%
Yes Yes No Yes No 84%
Yes No Yes No Yes 78%
No Yes No Yes Yes 78%
Yes Yes Yes No No 76%
No Yes Yes No Ye s 68%
Yes No Yes Yes No 68%
No No Yes Yes Ye s 58%
No Yes Yes Yes No 56%
2 out of 5 risk factors
Yes No No No Yes 72%
Yes Yes No No No 70%
No Yes No No Yes 61%
Yes No No Yes No 61%
No No No Ye s Yes 51%
Yes No Yes No No 48%
No Yes No Yes No 48%
No No Yes No Yes 38%
No Yes Yes No No 35%
No No Yes Yes No 27%
1 out of 5 risk factors
Yes No No No No 41%
No No No No Yes 31%
No Yes No No No 29%
No No No Ye s No 21%
No No Yes No No 14%
AFP, alpha-fetoprotein.
© 2022 The Author(s). Published by Wolters Kluwer Health, Inc. 7Baghdadi et al
especially in Asia. Our transplant center also uses the UCSF
downstaging protocol, which includes initial tumor burden
outside Milan criteria; this may not apply to all other trans-
plant centers. Lastly, regional variability should be consid-
ered, as the average waitlist period can impact the incidence
of HCC progression when longer waitlist time is associated
with higher dropout. Future prospective studies with a larger
sample size are needed to validate our results.
In conclusion, HCC progression and dropout from the
transplant list are associated with tumor size, the number of
lesions, capsule bulging, tumor margin, and AFP levels at base-
line. A risk model built from these variables can be used as
a predictive test for HCC progression within the transplant
waitlist. Overall, our ndings imply that HCC patients can be
risk-stratied from their baseline MR imaging; those identied
as high risk for progression and delisting may benet from
more aggressive treatment, which is now available for HCC.
REFERENCES
1. Howlader N, Noone AM, Krapcho M, et al. SEER Cancer Statistics
Review, 1975–2018. National Cancer Institute; 2021.
2. Mehta N, Fidelman N, Sarkar M, et al. Factors associated with out-
comes and response to therapy in patients with infiltrative hepatocel-
lular carcinoma. Clin Gastroenterol Hepatol. 2013;11:572–578.
3. Kim WR, Lake JR, Smith JM, et al. OPTN/SRTR 2013 Annual Data
Report: liver. Am J Transpl [Internet]. 2015;15 Suppl 2:1–28.
4. Yao FY, Fidelman N. Reassessing the boundaries of liver transplan-
tation for hepatocellular carcinoma: where do we stand with tumor
down-staging? Hepatology. 2016;63:1014–1025.
5. Graziadei I. Chemoembolization followed by liver transplantation
for hepatocellular carcinoma impedes tumor progression while
on the waiting list and leads to excellent outcome. Liver Transpl.
2003;9:557–563.
6. Mazzaferro V, Regalia E, Doci R, et al. Liver transplantation for the
treatment of small hepatocellular carcinomas in patients with cirrhosis.
N Engl J Med. 1996;334:693–699.
7. Yao FY, Ferrell L, Bass NM, et al. Liver transplantation for hepatocel-
lular carcinoma: expansion of the tumor size limits does not adversely
impact survival. Hepatology. 2001;33:1394–1403.
8. Lee HA, Cho EY, Kim TH, et al. Risk factors for dropout from the
liver transplant waiting list of hepatocellular carcinoma patients under
locoregional treatment. Transplant Proc. 2018;50:3521–3526.
9. Davis GL, Dempster J, Meler JD, et al. Hepatocellular carcinoma:
management of an increasingly common problem. Proc (Bayl Univ
Med Cent). 2008;21:266–280.
10. Bangaru S, Marrero JA, Singal AG. Review article: new therapeu-
tic interventions for advanced hepatocellular carcinoma. Aliment
Pharmacol Ther. 2020;51:78–89.
11. Finn RS, Qin S, Ikeda M, et al.; IMbrave150 Investigators. Atezolizumab
plus bevacizumab in unresectable hepatocellular carcinoma. N Engl J
Med. 2020;382:1894–1905.
12. El-Khoueiry AB, Sangro B, Yau T, et al. Nivolumab in patients with
advanced hepatocellular carcinoma (CheckMate 040): an open-label,
non-comparative, phase ½ dose escalation and expansion trial.
Lancet. 2017;389:2492–2502.
13. Kudo M, Finn RS, Qin S, et al. Lenvatinib versus sorafenib in first-line
treatment of patients with unresectable hepatocellular carcinoma: a
randomised phase 3 non-inferiority trial. Lancet. 2018;391:1163–1173.
14. Mehta N, Dodge JL, Hirose R, et al. Predictors of low risk for dropout
from the liver transplant waiting list for hepatocellular carcinoma in long
wait time regions: implications for organ allocation. Am J Transplant.
2019;19:2210–2218.
15. Toso C, Dupuis-Lozeron E, Majno P, et al. A model for dropout assess-
ment of candidates with or without hepatocellular carcinoma on a
common liver transplant waiting list. Hepatology. 2012;56:149–156.
16. Baghdadi A, Ghadimi M, Mirpour S, et al. Imaging neuroendocrine
tumors: characterizing the spectrum of radiographic findings. Surg
Oncol. 2021;37:101529.
17. Hazhirkarzar B, Tang H, Ghadimi M, et al. Predicting early necrosis of
colorectal liver metastases using volumetric enhancement on base-
line MRI and oil deposition on Post-cTACE unenhanced CT. Abdom
Radiol. 2021;46:4610–4618.
18. Ghadimi M, Pandey P, Rezvani Habibabadi R, et al. Role of volumet-
ric multiparametric MRI in distinguishing between intraductal papil-
lary mucinous neoplasms and serous cystadenoma. Abdom Radiol.
2021;46:1629–1639.
19. Shaghaghi M, AliyariG Hasabeh M, Ameli S, et al. Role of tumor mar-
gin and ADC change in defining the need for additional treatments
after the first TACE in patients with unresectable HCC. Eur J Radiol.
2020;133:109389.
20. Aliyari Ghasabeh M, Shaghaghi M, Pandey A, et al. Integrating base-
line MR imaging biomarkers into BCLC and CLIP improves overall sur-
vival prediction of patients with hepatocellular carcinoma (HCC). Eur
Radiol. 2021;31:1630–1641.
21. Pandey A, Pandey P, Ghasabeh MA, et al. Baseline volumetric mul-
tiparametric MRI: can it be used to predict survival in patients with
unresectable intrahepatic cholangiocarcinoma undergoing transcath-
eter arterial chemoembolization? Radiology. 2018;289:843–853.
22. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas
under two or more correlated receiver operating characteristic curves:
a nonparametric approach. Biometrics. 1988;44:837–845.
23. LeDell E, Petersen M, van der Laan M. Computationally efficient con-
fidence intervals for cross-validated area under the ROC curve esti-
mates. Electron J Stat. 2015;9:1583–1607.
24. Yao FY, Bass NM, Nikolai B, et al. Liver transplantation for hepato-
cellular carcinoma: analysis of survival according to the intention-
to-treat principle and dropout from the waiting list. Liver Transpl.
2002;8:873–883.
25. Freeman RB, Edwards EB, Harper AM. Waiting list removal rates
among patients with chronic and malignant liver diseases. Am J
Transplant. 2006;6:1416–1421.
26. Mehta N, Dodge JL, Roberts JP, et al. A novel waitlist dropout score
for hepatocellular carcinoma - identifying a threshold that predicts
worse post-transplant survival. J Hepatol. 2021;74:829–837.
27. Aitcheson G, Pillai A, Dahman B, et al. Recent advances in systemic
therapies for advanced hepatocellular carcinoma. Curr Hepatol Rep.
2021;20:23–33.
28. Li X, Wang Y, Ye X, Liang P. Locoregional combined with systemic
therapies for advanced hepatocellular carcinoma: an inevitable trend
of rapid development. Front Mol Biosci. 2021;8:1–11.
29. Kim HC, Chung JW, Lee W, et al. Recognizing extrahepatic collateral
vessels that supply hepatocellular carcinoma to avoid complications
of transcatheter arterial chemoembolization. Radiographics. 2005;25
Suppl 1:S25–S39.
30. Elsahhar A, Abdelwahab SM, Nasser HM, et al. Assessment of
the relationship between hepatocellular carcinoma location and its
response to transarterial chemoembolization. Egypt J Radiol Nucl
Med. 2021;52:220.
31. Server S, Sabet S, Yaghouti K, et al. Value of imaging findings in the
prediction of microvascular invasion in hepatocellular carcinoma.
Transplant Proc. 2019;51:2403–2407.
32. Ameli S, Shaghaghi M, Aliyari Ghasabeh M, et al. Role of base-
line volumetric functional MRI in predicting histopathologic grade
and patients’ survival in hepatocellular carcinoma. Eur Radiol.
2020;30:3748–3758.