European Journal of Cancer 82 (2017) 103e114

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.ejcancer.com

Original Research

Different patterns in the risk of newly developed fatty

liver and lipid changes with tamoxifen versus aromatase
inhibitors in postmenopausal women with early breast
cancer: A propensity scoreematched cohort study

Namki Hong a,b, Han Gyul Yoon a, Da Hea Seo a, Seho Park c,
Seung Il Kim c, Joo Hyuk Sohn d, Yumie Rhee a,*

a
Department of Internal Medicine, Endocrine Research Institute, Severance Hospital, Yonsei University College of Medicine,
Seoul 03722, South Korea
b
Graduate School, Yonsei University College of Medicine, Seoul 03722, South Korea
c
Department of Surgery, Severance Hospital, Yonsei University College of Medicine, Seoul 03722, South Korea
d
Division of Medical Oncology, Department of Internal Medicine, Yonsei Cancer Center, Yonsei University College of
Medicine, Seoul 03722, South Korea

Received 23 October 2016; received in revised form 26 January 2017; accepted 4 May 2017

KEYWORDS Abstract Background: Management of metabolic complications of long-term adjuvant endo-

Adjuvant endocrine crine therapy in early breast cancer remained an unmet need. We aimed to compare the effects
therapy; of tamoxifen (TMX) and aromatase inhibitors (AIs) on the risk of fatty liver in conjunction
Tamoxifen; with longitudinal changes in the serum lipid parameters.
Anastrozole; Methods: Among 1203 subjects who were taking adjuvant TMX or AI (anastrozole or letro-
Letrozole; zole) without fatty liver at baseline, those taking TMX or AI were 1:1 matched on the propen-
Hepatic steatosis; sity score. The primary outcome was newly developed fatty liver detected on annual liver
Triglyceride; ultrasonography.
High-density Results: Among 328 matched subjects (mean age 53.5 years, body mass index 22.9 kg/m2), 62
lipoprotein cholesterol cases of fatty liver in the TMX group and 41 cases in the AI group were detected in a total of
987.4 person-years. The incidence rate of fatty liver was higher in the TMX group than in the
AI group (128.7 versus 81.1 per 1000 person-years, P Z 0.021), particularly within the first 2
years of therapy. TMX was associated with an increased 5-year risk of newly developed fatty
liver (adjusted hazard ratio 1.61, P Z 0.030) compared with AI independent of obesity and
cholesterol level. Subjects who developed fatty liver had higher triglycerides (TGs) and

* Corresponding author: Department of Internal Medicine, Yonsei University College of Medicine, 50-1 Yonsei-ro, Seodaemun-gu, Seoul 03722,
South Korea. Fax: þ82 2 392 5548.
E-mail address: yumie@yuhs.ac (Y. Rhee).

http://dx.doi.org/10.1016/j.ejca.2017.05.002
0959-8049/ª 2017 Elsevier Ltd. All rights reserved.
104 N. Hong et al. / European Journal of Cancer 82 (2017) 103e114

lower high-density lipoprotein cholesterol (HDL-C) level at baseline than those without,
which was sustained during follow-up despite the serum cholesterolelowering effect of TMX.
Conclusions: TMX independently increased the 5-year risk of newly developed fatty liver
compared with AI in postmenopausal women with early breast cancer. Our findings suggest
the need for considering the risk of fatty liver as a different adverse event profile between
AI and TMX, particularly in patients with obesity, high TGs and low HDL-C.
ª 2017 Elsevier Ltd. All rights reserved.

1. Introduction cancer, along with identifying the independent pre-

dictors of incident fatty liver.
Breast cancer is the most common cancer in women,
with an annual incidence of more than 1.7 million
2. Methods
worldwide [1]. Adjuvant endocrine therapies such as
tamoxifen (TMX), aromatase inhibitors (AIs) or a
2.1. Study design and subjects
sequence of these agents are now recognised as the vital
strategy based on the improved survival among post-
We conducted a propensity scoreematched retrospec-
menopausal women with hormone receptorepositive
tive cohort study by using longitudinal data retrieved
breast cancer, which accounts for most breast cancer
from the electronic registry of a tertiary-level, university-
cases [2]. As longer durations of endocrine therapies (>5
affiliated single institution (Severance Hospital, Yonsei
years) with either ongoing TMX, AI or sequential
University College of Medicine, Seoul, Korea). Between
therapy are now being recommended, there has been
January 2006 and May 2015, a total of 5250 subjects
increasing attention to the impact of long-term endo-
received a diagnosis of hormone receptorepositive
crine treatments on metabolic consequences, including
breast cancer for the first time in their lifetime, and were
the development of fatty liver and derangement of lipid
started on adjuvant hormone therapies after the primary
metabolism [2e5]. Given the improved disease-free
surgery (Fig. 1). The diagnosis of breast cancer was
survival in postmenopausal women with early breast
defined by using relevant International Classification of
cancer, the proper management of metabolic risk in this
Diseases 10th revision (ICD-10) codes for breast cancer
population remains an unmet need.
(C50). Adjuvant endocrine therapies were defined as the
Along with the global epidemic of obesity, non-
presence of any prescription of 20-mg TMX daily or
alcoholic fatty liver disease (NAFLD) is the major cause
non-steroidal third-generation AI (1-mg anastrozole or
of liver diseases worldwide, which is linked to type 2
2.5-mg letrozole daily). The index date for study entry
diabetes, increased cardiovascular risk and kidney
was defined as the first prescription date of adjuvant
complications [6,7]. TMX was reported to be associated
endocrine therapies. The baseline data of study subjects
with hepatic steatosis and steatohepatitis either by
within 6 months before the index date, including de-
directly enhancing hepatic lipogenesis or by precipi-
mographics, cancer stage, pathology reports, liver ul-
tating NAFLD through the exacerbation of insulin
trasonography, comorbidities, alcohol intake, smoking,
resistance, central obesity and triglyceridemia [8,9].
medications including adjuvant chemotherapy and lab-
However, data on the long-term effects of TMX on the
oratory values, were obtained by using relevant ICD
development of fatty liver in postmenopausal women
codes and electronic medical records. The presence of
are sparse [10]. Despite the expanding use of AI as the
diabetes was defined as having relevant ICD-10 codes
upfront adjuvant endocrine therapy in postmenopausal
plus any prescriptions for diabetes medications or hae-
women, few data are available about the effect of these
moglobin A1c  6.5%. Exclusion criteria were as fol-
drugs on the incidence of fatty liver in comparison with
lows: no available liver ultrasonography data at
TMX [11]. Furthermore, the effect of TMX and AI on
baseline; metastatic breast cancer; poor Eastern Coop-
the longitudinal changes of serum lipid parameters in
erative Oncology Group performance score (ECOG-PS
conjunction with the development of fatty liver has not
3e4); uncertain cancer stage at index date; history of
been investigated.
liver cirrhosis, viral or alcoholic hepatitis, non-alcoholic
In this propensity scoreematched cohort study, we
steatohepatitis, hepatitis B or C infection and hyper- or
aimed to compare the effect of TMX and AI on the risk
hypothyroidism; current alcohol intake defined as any
of newly developed fatty liver and the longitudinal
alcohol intake within 1 year; presence of fatty liver at
change of serum lipid parameters during 5 years of
baseline and potential premenopause status defined as
upfront adjuvant endocrine therapy in postmenopausal
age at index date < 60 years plus baseline follicular
women with hormone receptorepositive early breast
stimulating hormone level < 25 mIU/mL [12]. To
 N. Hong et al. / European Journal of Cancer 82 (2017) 103e114
Fig. 1. Study flowchart of the propensity scoreematched cohort. ‘TMX’ indicates subjects treated with adjuvant 20-mg tamoxifen daily,
whereas ‘AI’ indicates those treated with aromatase inhibitors, either with 1-mg anastrozole or 2.5-mg letrozole daily.
compare the switching effect of TMX and AI on the risk
of fatty liver, subjects who received sequential adjuvant
endocrine therapies (switching from TMX to AI or vice
versa) within 5 years of follow-up were analysed as a
separate cohort. The subjects were followed for 5 years
or until the detection of newly developed fatty liver on
ultrasonography. This study was approved by the
Institutional Review Board of Severance Hospital,
Yonsei University (No. 4-2016-0525).
2.2. Measurements
At baseline and during the follow-up period, liver ul-
trasonography was performed in study subjects by
experienced radiologists in a single institution, on an
annual basis. Fatty liver was detected and graded
semiquantitatively (mild, moderate or severe) on the
basis of characteristics of diffuse hyperechogenicity of
the liver compared with the cortex of the right kidney,
poor visualisation of intrahepatic vessel walls and ul-
trasound beam attenuation [13]. The development of
any degree of fatty liver during the follow-up period was
defined as the primary end-point. The secondary end-
point was moderate to severe fatty liver defined as a
composite of (1) fatty liver graded as moderate or severe
by ultrasonography and (2) any degree of fatty liver plus
elevation of aspartate aminotransferase (AST) or
105
alanine aminotransferase (ALT) above the upper limit
of normal at the time of fatty liver detection. Fasting
blood samples were obtained through antecubital veni-
puncture at the outpatient clinic, at baseline and during
follow-up. The serum levels of total cholesterol (TC),
high-density lipoprotein cholesterol (HDL-C), triglyc-
eride (TG), glucose, alkaline phosphatase (ALP),
AST and ALT were measured by using a Hitachi 7600
autoanalyzer (Hitachi Ltd., Tokyo, Japan) in a central
laboratory. Low-density lipoprotein cholesterol (LDL-
C) was calculated by using Friedewald’s equation: LDL-
C (mg/dL) Z TC (mg/dL) e HDL-C (mg/dL) e TG
(mg/dL)/5. The reference ranges for the liver enzymes
were as follows: AST, 13.0e34.0 U/L; ALT, 5.0e46.0 U/
L; and ALP, 38.0e115.0 U/L. The hepatic steatosis
index (HSI) at baseline was calculated as follows:
HSI Z 8  AST/ALT ratio þ body mass index (BMI)
(þ2, if female; þ2, if with diabetes) [14]. The homoeo-
static model assessment of insulin resistance (HOMA-
IR) was calculated in a population subset with available
fasting insulin data as follows: fasting insulin level (mU/
mL)  fasting plasma glucose level (mmol/L)/22.5 [15].
AST-to-platelet ratio index [APRI Z AST (U/L)/upper
normal of AST/platelet count (109/L)  100] of subjects
at baseline and at the time of fatty liver detection was
calculated to assess the probability of fibrosis along with
sonographic findings [16]. Cut-offs for low and high
106 N. Hong et al. / European Journal of Cancer 82 (2017) 103e114
probability of advanced fibrosis was set at 0.5 and 1.5 as
previously published [16]. Significant progression of
fibrosis was defined using arbitrary cutoff as the highest
quartile of delta of APRI (>0.14; median 0.07 with
interquartile range [IQR] 0.01e0.14).
2.3. Statistical analysis
Data are presented as mean  SD, median [IQR] or as
numbers (percentages). Independent t-test and chi-
square test were used to compare continuous and cate-
gorical variables, as appropriate. The propensity score
for being allocated to either the TMX group or the AI
group was estimated by using a logistic regression model
in subjects those who received TMX or AI as upfront
adjuvant therapy without switching (Fig. 1). Potential
risk factors related to outcome (incident fatty liver) and
confounding variables associated with both treatment
status (AI or TMX) and outcome were included as
propensity score covariates on the basis of literature
review and prior clinical knowledge, as follows: age,
BMI, cancer stage, ECOG-PS, smoking status, diabetes,
hypertension, duration of adjuvant endocrine therapy,
systolic blood pressure, diastolic blood pressure, adju-
vant chemotherapy regimens (doxorubicin, docetaxel,
paclitaxel and cyclophosphamide), adjuvant radio-
therapy, medications (statins, angiotensin inhibitors,
beta-blockers and antiplatelets) and laboratory values
(AST, ALT, ALP, total bilirubin, albumin, TC, HDL-C,
TG, LDL-C and glucose) [17e20]. To match subjects
between the TMX and AI groups, a nearest-neighbour
algorithm on a 1:1 basis without replacement within a
calliper of 0.25  standard deviation of log-transformed
propensity score was used [21]. Covariate balance was
checked by using the absolute standardised mean dif-
ference with a threshold of 0.2 to determine substantial
imbalance [21]. The time to newly developed fatty liver
between matched groups was compared by plotting
KaplaneMeier curves with a log-rank test. Multivariate
cox proportional hazard models for fatty liverefree
survival were created to identify the independent risk
factors for newly developed fatty liver. The changes of
mean serum lipid parameters between the TMX and AI
groups over time were compared by using linear mixed-
effects regression models for longitudinal repeated
measures with an autoregressive covariance pattern. The
models included the treatment group, time, age, BMI,
diabetes, use of statins, adjuvant chemotherapy as fixed
predictor variables and the group  time interactions.
Random intercepts for study subjects were included to
account for the dependence of repeated measures. No
variables included in the models violated the propor-
tional hazard assumption. In subjects those who
received sequential therapy during follow-up, multivar-
iate cox models including treatment status as time-
varying covariate were built to assess the effect of
switching therapy (from TMX to AI or vice versa) on
the risk of fatty liver. A two-sided P-value of <0.05 was
considered significant. All statistical analyses were per-
formed with STATA 14.0 (Stata Corp, College Station,
TX, USA).
3. Results
3.1. Baseline characteristics of the study subjects
The baseline characteristics of 1203 subjects in the un-
matched cohort are presented in Table A1. Compared
with those receiving TMX as adjuvant endocrine ther-
apy, those in the AI group were more likely to be older
and have higher BMI, higher prevalence of diabetes or
hypertension, previous medication use including statins,
higher fasting glucose level and longer duration of
adjuvant endocrine therapies. After matching based on
the propensity score, a total of 328 subjects remained in
the matched cohort and the baseline characteristics were
well balanced without significant differences between the
TMX and AI groups (Table 1; Fig A.1). Among the AI
group, 76 subjects received anastrozole (46.3%) and 88
were taking letrozole (53.7%). The mean age of the
matched subjects was 53.5  7.7 years with a mean BMI
of 22.9  2.8 kg/m2. About 50% of the subjects had
stage 1 cancer, and the proportions of those who
received adjuvant chemotherapy or radiotherapy did not
differ significantly between the two groups. The mean
serum AST, ALT, albumin, TC, TG and HDL-C levels
were within the reference ranges in both groups. No
previous use of thiazolidinedione or vitamin E was
found in study subjects at baseline. The median duration
of adjuvant endocrine therapies was 3.5 years [IQR
2.0e5.0].
3.2. Incidence of fatty liver between the TMX and AI
groups during follow-up
In a total of 987.4 person-years in the matched cohort,
103 cases of newly developed fatty liver were detected
(62 cases in the TMX group and 41 cases in the AI
group). Subjects who received TMX had a significantly
higher incidence rate of fatty liver than those in the AI
group (128.7 versus 81.1 per 1000 person-years,
P Z 0.021). The proportion of incident fatty liver be-
tween subjects taking anastrozole and those taking
letrozole within the AI group did not differ significantly
(22 of 76 versus 19 of 88 cases, P Z 0.278). In
KaplaneMeier plots, a steeper drop in fatty liverefree
probability was observed in the TMX group than in the
AI group, particularly within the first 2 years of therapy
(log-rank P Z 0.017, Fig. 2A). Most cases of moderate
to severe fatty liver as secondary end-point were detec-
ted in the TMX group (n Z 27, 16.7% versus n Z 2,
1.2%, P < 0.001), with a significantly higher incidence
rate than in the AI group (48.1 versus 3.2 per 1000
 N. Hong et al. / European Journal of Cancer 82 (2017) 103e114 107

Table 1
Baseline characteristics of study participants in the propensity score-matched cohort.
Variables Tamoxifen (N Z 164) Aromatase inhibitors (N Z 164) P-value
Age, year 53.0  8.8 54.1  6.5 0.164
BMI, kg/m2 22.9  2.9 22.9  2.8 0.915
ECOG-PS 0.975
0 151 (92.1) 150 (91.5)
1 10 (6.1) 11 (6.7)
2 3 (1.8) 3 (1.8)
Cancer stage 0.770
0 14 (8.5) 10 (6.1)
1 86 (52.4) 88 (53.7)
2 55 (33.5) 54 (32.9)
3A 9 (5.6) 12 (7.3)
Histopathologic type 0.738
Invasive ductal carcinoma 135 (82.3) 139 (84.7)
Ductal carcinoma in situ 14 (8.5) 10 (6.1)
Invasive lobular carcinoma 7 (4.3) 9 (5.5)
Othera 8 (4.9) 6 (3.7)
Hormone receptor status
ER positive 160 (97.6) 158 (96.3) 0.521
PR positive 126 (76.8) 99 (60.4) 0.001
HER-2 positiveb 23 (14.0) 21 (12.8) 0.746
Alcohol
Never 155 (94.5) 158 (96.3) 0.428
Ex-drinker 9 (5.5) 6 (3.7)
Smoking
Never 159 (97.0) 158 (96.3) 0.930
Ex-smoker 2 (1.2) 2 (1.2)
Current 3 (1.8) 4 (2.5)
Diabetes 15 (9.2) 23 (14.0) 0.168
Hypertension 17 (10.4) 17 (10.4) 1.000
Chemotherapy
Adriamycin 104 (63.4) 96 (58.5) 0.365
Docetaxel 36 (21.9) 35 (21.3) 0.893
Paclitaxel 20 (12.2) 15 (9.2) 0.371
Cyclophosphamide 97 (59.2) 94 (57.3) 0.737
Radiation therapy 113 (68.9) 104 (63.4) 0.294
Medications
Statins 4 (2.4) 6 (3.7) 0.521
Metformin 2 (1.2) 4 (2.4) 0.410
Angiotensin blockers 6 (3.7) 9 (5.5) 0.428
Adjuvant endocrine therapy duration, days 3.5 [2.2e5.0] 3.5 [1.9e5.0] 0.224
Laboratory values
AST, IU/L 20.7  6.5 21.8  9.2 0.246
ALT, IU/L 19.1  11.4 20.1  12.0 0.423
ALT/AST ratio 0.9  0.3 0.9  0.2 0.774
ALP, IU/L 58.3  15.8 54.3  20.2 0.460
Total bilirubin, mg/dL 0.6  0.3 0.5  0.2 0.464
Albumin, g/dL 4.3  0.4 4.2  0.5 0.111
Total cholesterol, mg/dL 202.5  35.3 200.0  41.3 0.487
Triglyceride, mg/dL 133.9  132.1 129.5  115.1 0.748
HDL-cholesterol, mg/dL 52.1  11.9 52.4  12.4 0.817
LDL-cholesterol, mg/dL 125.8  33.8 120.9  41.7 0.243
Fasting glucose, mg/dL 101.8  25.8 102.5  18.9 0.773
Hepatic steatosis index 35.2  4.9 35.0  4.1 0.559
Data are presented as mean  standard deviation, median [interquartile range] or number (%).
Abbreviations: BMI, body mass index; ECOG-PS, Eastern Cooperative Oncology Group performance score; BP, blood pressure; AST, aspartate
aminotransferase; ALT, alanine aminotransferase; ALP, alkaline phosphatase; HDL, high-density lipoprotein; LDL, low-density lipoprotein; ER,
oestrogen receptor; PR, progesterone receptor; IHC, immunohistochemistry.
a
Other histopathologic types include tubular, papillary, micropapillary, mucinous, Paget’s disease and cribriform.
b
HER-2 positive is defined as IHC 3þ or IHC 2þ and Fluorescence In Situ Hybridization (FISH) positive.
108 N. Hong et al. / European Journal of Cancer 82 (2017) 103e114
Fig. 2. KaplaneMeier curves for event-free probability of (A) any
newly developed fatty liver detected on liver ultrasonography and
(B) the composite of moderate to severe hepatic steatosis on liver
ultrasonography and steatohepatitis defined as any grade of fatty
liver with elevated serum aspartate aminotransferase or alanine
aminotransferase level above the reference range. Compared with
AI, the incidence rate of newly developed fatty liver increased
steeply in the TMX group within the first 2 years and remained
similar to AI thereafter. Most cases of moderate to severe fatty
liver occurred in the TMX group. Abbreviations: AI, aromatase
inhibitors; TMX, tamoxifen.
person-years, P < 0.001; Fig. 2B). Fatty liver cases in the
AI group were mostly attributed to mild-grade fatty
liver without AST or ALT elevation. KaplaneMeier
plots in unmatched cohort revealed similar pattern (Fig
A.2).
3.3. Probability of fibrosis in subjects who developed fatty
liver disease
Among subjects who developed fatty liver (n Z 103),
APRI at baseline were within the range of low proba-
bility of fibrosis (APRI < 0.5) except one in TMX group
with intermediate probability (APRI 0.5e1.5). Mean
APRI at baseline did not differ significantly between
TMX and AI groups (0.22  0.11 versus 0.22  0.08,
P Z 0.779), whereas APRI at the time point of fatty
liver detection was increased in both TMX (delta of
APRI 0.11  0.21, P < 0.001) and AI groups
(0.05  0.10, P Z 0.006). TMX use was associated with
significant progression of fibrosis (delta of APRI > 0.14)
compared with AI use (n Z 20, 32.3% versus n Z 6,
14.6%, P Z 0.044), which resulted in the higher pro-
portion of subjects in the TMX group having interme-
diate fibrosis probability combined with fatty liver than
in the AI group (n Z 6, 9.7% versus n Z 0, 0.0%,
P Z 0.040). Subjects who developed moderate to severe
fatty liver disease were more likely to have significant
progression of fibrosis compared with those who
developed mild fatty liver (44.8% versus 17.6%,
P Z 0.004). All subjects with intermediate probability of
fibrosis at the time of fatty liver detection belonged to
the moderate to severe fatty liver disease group.
3.4. Changes in lipid parameters over time
When the longitudinal changes of serum lipid parame-
ters were compared between the TMX and AI groups in
the matched cohort, a significant decrease in the TC
level over time was observed in subjects who received
TMX, in contrast to the unchanged TC level in the AI
group (P for interaction < 0.001, Fig. 3A), which
resulted in a significantly lower TC level in the TMX
group than in the AI group at any time point after 1 year
of therapy. The HDL-C level increased gradually over
time in both groups; however, no significant differences
were observed between the TMX and AI groups at any
time point. Despite the significant decrease in the TC
level in the TMX group, the TG level in this group
remained unchanged, similar to the AI group. When
changes of lipid parameters were plotted as grouped by
the development of fatty liver during follow-up
(Fig. 3B), the patterns of change in TC, HDL-C,
TG and LDL-C did not differ significantly between
subjects who developed fatty liver during follow-up and
those who did not, within each treatment group. How-
ever, in both the TMX and AI groups, a significantly
higher TG level and a lower HDL-C level at baseline
were observed in the fatty liver group compared with
those without fatty liver. The gap in the TG and HDL-C
levels at baseline was attenuated but maintained statis-
tical significance over time, independent of the treatment
group. In contrast with the decrease in TG level over
time in the AI fatty liver group, the TG level remained
unchanged in the TMX fatty liver group despite the
decrease in TC. The follow-up BMI at a median of 2.5
years [IQR 1.5e3.5] did not differ significantly from the
baseline BMI in both the TMX (22.8  2.8 to
22.9  2.7 kg/m2, P Z 0.937) and AI (22.9  2.8 to
22.6  2.8 kg/m2, P Z 0.057) groups. BMI was main-
tained without significant changes in subjects who
developed fatty liver (24.1  2.6 versus 24.1  2.5 kg/m2,
P Z 0.796), whereas BMI in those without incident fatty
liver slightly decreased from 22.3  2.7 to 22.1  2.6 kg/
m2 with marginal statistical significance (P Z 0.051).
 N. Hong et al. / European Journal of Cancer 82 (2017) 103e114
3.5. Independent predictors of newly developed fatty liver
TMX use versus AI use (hazard ratio [HR] 1.61,
P Z 0.018), younger age, higher baseline BMI, ALT/
AST ratio, TG level and lower HDL-C level were
significantly associated with an increased risk of newly
developed fatty liver during follow-up in univariate cox
proportional hazard models (Table 2). The association
of the presence of diabetes, fasting glucose level, cancer
stage, adjuvant chemotherapy and changes of BMI
during follow-up with incident fatty liver did not show
statistical significance. In the multivariate cox model,
adjuvant TMX therapy increased the risk of newly
developed fatty liver by 61% compared with AI treat-
ment (adjusted HR 1.61, P Z 0.030) independent of
potent risk factors including age, BMI, HDL-C, TG and
ALT/AST ratio, which also remained as significant
predictors in the multivariate model. Subjects in TMX
group were also strongly associated with elevated risk of
severe fatty liver as secondary composite end-point
compared with AI users (adjusted HR 17.8, P < 0.001).
TMX use was also an independent predictor for fatty
liver (adjusted HR 1.42, P Z 0.026) in 1203 subjects in
the unmatched cohort when the same multivariate
model was applied (Table A.2). The independent asso-
ciation of TMX with incident fatty liver was not atten-
uated even when the analyses were confined to the
subset of subjects who continued adjuvant endocrine
therapy for at least 3 years (TMX, n Z 103 versus AI,
n Z 95; adjusted HR 1.86, P Z 0.010) and for 5 years or
more (TMX, n Z 41 versus AI, n Z 41; adjusted HR
2.56, P Z 0.013). In a subset of matched subjects with
available fasting insulin level (n Z 260 of 368, 79.3%), a
HOMA-IR above the median value (1.40) was signif-
icantly associated with an increased risk of fatty liver
(HR 1.97, P Z 0.002). When HOMA-IR was entered
into the multivariate model, the predictive value of
TMX use versus AI use remained robust (adjusted HR
1.70, 95% confidence interval [CI] 1.05e2.75,
P Z 0.030), whereas the statistical significance of HDL-
C (adjusted HR 0.97, P Z 0.059) and TG (adjusted HR
1.09, P Z 0.296) was attenuated.
3.6. Progress of fatty liver after initial detection
Among 103 subjects who developed fatty liver, data of
follow-up ultrasonography during study period was
available in 83 subjects (80.6%). The median interval
between the first detection of fatty liver and follow-up
was 2.5 years [IQR 1.5e3.0]. Adjuvant endocrine ther-
apy was continued after first detection until the follow-
up ultrasonography in 71 subjects (85.5%). In 12 sub-
jects, TMX or AI was discontinued at median 187
d [31e315] before follow-up ultrasonography. Overall,
TMX use was associated with persistent or progression
to fatty liver with moderate to severe degree compared
with AI use (n Z 22/52, 42.3% versus n Z 3/31, 9.7%,
109
P Z 0.002). In a subgroup of subjects who discontinued
TMX or AI before follow-up sonography (n Z 12), 3 of
6 subjects treated with TMX revealed persistent (n Z 1)
or progression (n Z 2) to moderate to severe degree
fatty liver even after treatment discontinuation, whereas
fatty liver remained as mild degree (n Z 4) or resolved
(n Z 2) in all the 6 subjects in the AI group. In multi-
variate logistic regression analysis, TMX use (OR 7.12,
95% CI 1.81e28.00, P Z 0.005) and higher baseline
BMI (OR 1.35, 95% CI 1.07e1.69, P Z 0.009) were the
significant risk factors for persistent or progression to
moderate to severe fatty liver after adjustment for age,
TG, HDL-C and ALT level.
3.7. Altered risk of incident fatty liver by sequential
therapy
Differences in the risk of fatty liver according to
switching therapy was assessed in a separate cohort of
255 subjects (TMX followed by AI [T-A], n Z 203; AI
followed by TMX [A-T], n Z 52). Compared to A-T
group, T-A group showed younger age (46.9  6.4
versus 52.4  10.8 years, P < 0.001), similar BMI
(22.1  2.6 versus 22.4  2.6 kg/m2, P Z 0.438) and
more favourable lipid profile (TG 113.7  62.2 versus
153.5  171.2 mg/dL, P Z 0.008; HDL-C 55.0  11.5
versus 50.6  11.6 mg/dL, P Z 0.018) at baseline. Me-
dian duration of endocrine therapy was 5.3 (3.6e7.3)
years (TMX 3.5 [1.5e5.2] followed by AI 1.9 [0.7e3.5])
for T-A group and 2.5 (1.3e4.6) years (AI 1.1 [0.3e2.4]
followed by TMX 0.8 [0.5e2.0]) for A-T group,
respectively. A total of 91 newly developed fatty liver
cases were detected (80.5 and 53.8 cases per 1000 person-
years for T-A and A-T groups, respectively, P Z 0.224).
When treatment status was entered into Cox model as
time-varying covariate, switching TMX to AI was
associated with reduced risk of fatty liver (HR 0.46, 95%
CI 0.25e0.86, P Z 0.016) in the T-A group and the
association was robust after adjustment for age, BMI,
TG, HDL-C and ALT level in multivariate model
(adjusted HR 0.44, 95% CI 0.22e0.87, P Z 0.018; Table
A.3.), whereas no significant change in risk of fatty liver
by switching AI to TMX was found (HR 0.68, 95% CI
0.16e2.80, P Z 0.595) in the A-T group.
4. Discussion
In this propensity scoreematched cohort study, we
found that adjuvant TMX treatment in postmenopausal
women with hormone receptorepositive early breast
cancer was significantly associated with an increased risk
of newly developed fatty liver, particularly with mod-
erate to severe fatty liver, compared with adjuvant AI
during 5 years of follow-up. The association of TMX
use and incident fatty liver was independent of con-
ventional risk factors for fatty liver, including BMI, TG,
110 N. Hong et al. / European Journal of Cancer 82 (2017) 103e114

Fig. 3. Longitudinal changes of serum lipid parameters (A) by treatment groups (TMX or AI) and (B) by the development of fatty liver
during follow-up within each treatment group. Data were analysed by using linear mixed models, including treatment group, time, age,
body mass index, diabetes, use of statins, adjuvant chemotherapy and group  time interaction as fixed predictor variables, as well as
random intercepts for study subjects with an autoregressive covariate pattern. In A, * indicates P < 0.05 versus baseline within treatment
 N. Hong et al. / European Journal of Cancer 82 (2017) 103e114 111

Table 2
Independent predictors for newly developed fatty liver in postmenopausal women with early breast cancer on adjuvant endocrine therapies.
Univariate Multivariate
Variables HR (95% CI) P-value HR (95% CI) P-value
TMX use (versus AI use) 1.61 (1.08e2.39) 0.018 1.61 (1.04e2.48) 0.030
Age 0.94 (0.91e0.98) <0.001 0.91 (0.86e0.95) <0.001
BMI 1.14 (1.09e1.21) <0.001 1.18 (1.10e1.25) <0.001
Cancer stage (versus stage 0)
Stage 1 1.29 (0.58e2.85) 0.520 0.60 (0.23e1.56) 0.303
Stage 2 1.45 (0.65e3.25) 0.359 0.59 (0.22e1.56) 0.289
Stage 3A 0.75 (0.24e2.37) 0.631 0.34 (0.09e1.24) 0.104
ECOG-PS (versus ECOG 0)
ECOG 1 0.83 (0.38e1.78) 0.626 0.88 (0.38e2.02) 0.775
ECOG 2 0.65 (0.16e2.63) 0.543 1.13 (0.26e4.87) 0.868
Current smoking 1.06 (0.26e4.32) 0.928 1.01 (0.42e2.38) 0.991
Ex-drinker (versus non-drinker) 1.45 (0.67e3.12) 0.340 1.14 (0.50e2.62) 0.744
Diabetes 0.75 (0.38e1.49) 0.421 0.66 (0.31e1.42) 0.288
Hypertension 0.79 (0.41e1.53) 0.496 1.24 (0.56e2.76) 0.583
Adjuvant chemotherapy 1.33 (0.86e2.07) 0.197 0.69 (0.38e1.25) 0.231
Statins 0.30 (0.41e2.14) 0.229 0.29 (0.03e2.40) 0.255
Angiotensin blockers 0.47 (0.15e1.49) 0.202 0.88 (0.22e3.47) 0.857
ALT/AST ratio 3.45 (1.84e6.33) <0.001 2.22 (1.10e4.49) 0.026
Albumin (per 1 g/dL increase) 0.77 (0.51e1.16) 0.221 0.71 (0.44e1.15) 0.169
HDL-cholesterol (per 1 mg/dL increase) 0.97 (0.95e0.98) 0.001 0.97 (0.95e0.99) 0.013
Triglyceride (per 1 sd increase) 1.25 (1.12e1.41) <0.001 1.15 (1.01e1.31) 0.036
Fasting glucose (per 1 sd increase) 0.99 (0.82e1.22) 0.989 1.26 (0.99e1.62) 0.056
Values with statistical significance are printed in bold.
Abbreviations: HR, hazard ratio; TMX, tamoxifen; AI, aromatase inhibitors; BMI, body mass index; ECOG-PS, Eastern Cooperative Oncology
Group performance score; AST, aspartate aminotransferase; ALT, alanine aminotransferase; HDL, high-density lipoprotein; CI, confidence in-
terval; sd, standard deviation.

HDL-C level and insulin resistance index. Although
TMX significantly decreased the TC level compared
with AI, and both TMX and AI modestly increased the
HDL-C level, subjects with incident fatty liver had
higher baseline serum TG and lower HDL-C level than
those without fatty liver, and the significant gap in TG
and HDL-C remained over time regardless of the effects
of TMX and AI on the lipid profiles.
Previous studies reported the association of TMX use
with several forms of drug-induced liver injury,
including the development of hepatic steatosis and
steatohepatitis [8,10,22]. After beginning TMX, fatty
liver detected on ultrasonography or computed tomog-
raphy appeared as early as 3 months and within the first
2 years of treatment in most cases [10,22]. The cumu-
lative incidence of fatty liver among subjects who
received TMX varied from 32.6% to 52.6% during 3e5
years of follow-up [10,22]. Our result was consistent
with previous findings showing that the cumulative
incidence of fatty liver in the TMX group (37.8%) was
within the reported range. Moreover, the risk of newly
developed fatty liver steeply increased within the first 2
years in the TMX group, similar to previous studies.
However, data on the comparison of the effect of AI on
fatty liver with that of TMX was not available in most
of the previous studies. In a prospective, randomised
trial, the cumulative incidence of fatty liver after 3 years
was significantly higher in the TMX arm than in the
anastrozole arm (41.1% versus 14.6%), similar to our
findings [11]. Despite the well-known differential effects
of TMX and AI on lipid profiles, data on serum lipid
parameters and insulin resistance, which could signifi-
cantly affect the development of fatty liver, particularly
in postmenopausal women, were not available in the
study [23,24]. In line with previous findings, we observed
that the mean serum TC level remained stable in the AI
group over time, whereas the serum TC level initially
decreased in the TMX group and then remained stable
thereafter [25]. The serum HDL-C level was modestly
increased in both treatment groups in our study. Despite
the observed effects of TMX and AI on the lipid pro-
files, subjects with incident fatty liver had higher TG and
lower HDL-C level than those without fatty liver, which
persisted during follow-up independent of the treatment
group. Taken together, our findings confirmed and
extended the findings of previous works by comparing

group, whereas y represents P < 0.05 versus AI. In B, * indicates P < 0.05 versus baseline over time and y represents P < 0.05 versus no
fatty liver group within each treatment regimen. Despite the significant decrease of TC level within the first 2 years of TMX treatment
compared with AI, the higher baseline TG level and lower HDL-C level in subjects with fatty liver compared with those without fatty liver
were not attenuated and remained significant over time. Abbreviations: AI, aromatase inhibitors; TMX, tamoxifen; HDL, high-density
lipoprotein; LDL, low-density lipoprotein.
112 N. Hong et al. / European Journal of Cancer 82 (2017) 103e114
the effect of upfront adjuvant TMX and AI on the
development of fatty liver during 5 years of follow-up, in
conjunction with the analysis of longitudinal changes of
serum lipid parameters in postmenopausal women with
early breast cancer.
NAFLD has been linked to increased cardiovascular
risk and long-term mortality in general population [26].
In particular, biopsy-proven fibrosis or higher proba-
bility of fibrosis estimated by non-invasive measure-
ments was the strongest predictor for mortality, with
cardiovascular disease being the main cause of death
[27,28]. Although the clinical significance of lone hepatic
steatosis diagnosed by ultrasonography without eleva-
tion of liver enzyme or evidences of fibrosis remained
questionable, cohort studies using paired biopsies
revealed that fibrosis progression occurred in both pa-
tients with steatohepatitis as well as hepatic steatosis,
suggesting the possibility of progression from simple
steatosis to steatohepatitis and clinically significant
fibrosis [29,30]. In our study, TMX use was strongly
related both to hepatic steatosis as well as development
of moderate to severe fatty liver combined with elevated
liver enzyme and higher probability of fibrosis estimated
by APRI, compared with AI use. In follow-up ultraso-
nography after development of fatty liver, TMX use was
a significant risk factor for persistent or progression to
moderate to severe fatty liver disease, along with high
BMI. Given the long-term clinical impact of NAFLD,
our findings may support the need for monitoring
development and progression of fatty liver during TMX
therapy, particularly in metabolically-prone subjects
with high BMI, although the effect of NAFLD during
adjuvant endocrine therapy on cancer-related outcomes
need further investigation.
Both direct and indirect mechanisms of TMX-
induced hepatosteatosis and steatohepatitis have been
suggested [8]. TMX is known to aggravate the predis-
position to NAFLD, potentially by increasing the
circulating TG level and promoting insulin resistance,
particularly in the obese population [9,31,32]. Our
findings support this notion that high BMI, serum TG
level and low HDL-C level were independently associ-
ated with the development of fatty liver in post-
menopausal women receiving adjuvant endocrine
therapy. Although a decrease in serum TC level was
observed in the TMX group as compared with the AI
group, the TG level remained high in subjects who
developed fatty liver in the TMX group, in contrast to
the gradual decreasing trend of TG level observed in
subjects in the AI group with incident fatty liver, sug-
gesting the possibility of TG dysregulation in
metabolically-prone subjects treated with TMX. A
direct hepatotoxicity of TMX was proposed, including
enhancement of hepatic de novo lipogenesis, inhibition
of mitochondrial fatty acid oxidation and decreased
incorporation of free fatty acids into very LDL (VLDL)
or attenuated secretion of VLDL [33e35]. On the basis
of these evidences, the robustness of the independent
predictive value of TMX versus AI use, even after
adjustment for metabolic profiles including the insulin
resistance index, may suggest the possible direct effects
of TMX on the development of fatty liver, which poses
additional risks compared with AI.
The cumulative incidence of newly developed fatty
liver in the AI group (25.0%) was relatively higher than
the previously reported incidence (ranging from 12.9%
to 14.6%); however, direct comparison with previous
findings need to be performed with caution owing to the
difference in the study design, such as a population with
younger age, subjects including both premenopausal
and postmenopausal women and use of different imag-
ing modalities for detecting fatty liver [10,11]. Given
that the incidence rate of fatty liver in the AI group in
our data (81.1 per 1000 person-years) were within the
reported range of NAFLD incidence rate in the general
population from community-based studies conducted in
Asia (52.3 and 91.0 per 1000 person-years), it is
conceivable that AI might have a neutral risk for fatty
liver, at least compared with TMX [6]. Furthermore,
switching TMX to AI was associated with reduced risk
of fatty liver in our study, which remained robust after
adjustment for potent confounders. Although we did
not find any significant differences in the risk of fatty
liver by switching AI to TMX, this result might be
attributed to the relatively small sample size of the AI to
TMX group compared with the TMX to AI group and
the differences in baseline metabolic profiles. Based on
these findings, it is conceivable that switching TMX to
AI in high-risk subjects might be one of the potential
strategies to prevent development of fatty liver during
adjuvant endocrine therapy. However, our study could
not draw any clear conclusion about whether AI treat-
ment itself might be neutral, beneficial or risky for the
development of fatty liver, owing to a lack of a control
group not receiving any adjuvant endocrine therapy.
Longitudinal studies comparing the effect of TMX,
AI or sequential therapy on the risk of fatty liver with
controls not receiving adjuvant hormone therapy would
be needed to clarify the association.
Recent epidemiological studies suggested possible
protective effect of modest alcohol consumption on
prevalence of fatty liver compared to complete absti-
nence [36,37]. Since alcohol consumption status was
initially collected as categorical variable (never; non-
drinker during past year and current drinker), data
regarding amount, type and frequency of alcohol intake
were not available in this study. Given the expected
heterogeneity in alcohol consumption pattern in current
drinkers, rigorous exclusion of current alcohol drinkers
was inevitable to reduce potential confounding.
Although previous alcohol intake was not a significant
predictor of incident fatty liver compared to lifetime
non-drinking in this study, further studies investigating
potential effect modification of the association between
 N. Hong et al. / European Journal of Cancer 82 (2017) 103e114
adjuvant endocrine therapy and the risk of fatty liver by
modest alcohol consumption might be helpful to extend
our finding to general population.
Our study is limited by possible residual confounding
factors despite the propensity score matching and sta-
tistical adjustment. The duration of adjuvant endocrine
therapy was suboptimal, although a high rate of early
discontinuation and non-adherence to adjuvant endo-
crine therapy in early-stage breast cancer has been well
recognised [38]. However, consistent results were
observed in our study when the subjects were confined
to those who took adjuvant endocrine therapy for the
entire duration. In this study, liver ultrasonography with
a semiquantitative grading system was used to define
newly developed fatty liver. Although ultrasonography
is known to be operator dependent, previous studies
reported the high sensitivity and specificity of ultraso-
nography (>90%) for the diagnosis of fatty liver,
particularly in patients with at least 20% steatosis [39].
Liver biopsy might provide further information on
characterising the histopathology of fatty liver, although
liver biopsy was not routinely available in this study
based on potential procedure-related morbidity [17].
Only anastrozole and letrozole as non-steroidal AI were
analysed in this study.
5. Conclusions
In conclusion, adjuvant TMX therapy was indepen-
dently associated with the increased risk of newly
developed fatty liver compared with AI therapy in
postmenopausal women with early breast cancer. Given
the potential risks of fatty liver on cardiovascular and
liver morbidity, the drug adherence of patients and the
absence of any single prognostic indicator that identifies
subjects who would benefit selectively from AI or TMX,
the different risk profile of fatty liver between AI and
TMX can be considered an important adverseeeffect
profile when initiating or switching adjuvant endocrine
therapy, particularly in postmenopausal women with
predisposing factors to NAFLD, including obesity, high
TG level and low HDL-C level.
Conflict of interest statement
None declared.
Funding
Not applicable.
Ethics approval and consent to participate
This study was approved by the Institutional Review
Board of Severance Hospital, Yonsei University (No. 4-
2016-0525).
113
Consent for publication
Not applicable.
Availability of data and material
The datasets during and/or analysed during the current
study are available from the corresponding author on
reasonable request.
Authors’ contributions
All authors had full access to all of the data in the study
and can take responsibility for the integrity of the data
and the accuracy of the data analysis; NH, HGY and
YR designed the study and wrote the protocol; HGY,
DHS, SP, SIK and JHS provided the data; NH under-
took data analysis; NH wrote the original draft of the
manuscript; SP, SIK, JHS and YR revised the final
manuscript and YR supervised data collection and
synthesis and is a guarantor.
Acknowledgements
The authors are grateful to Ji Soo Park (Division of
Medical Oncology, Department of Internal Medicine,
Yonsei Cancer Center, Yonsei University College of
Medicine, Seoul 03722, Korea) for sharing the clinical
experiences and providing thoughtful discussions about
long-term supportive care in breast cancer survivors.
Appendix A. Supplementary data
Supplementary data related to this article can be found
at http://dx.doi.org/10.1016/j.ejca.2017.05.002.
References
[1] Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D.
Global cancer statistics. CA Cancer J Clin 2011;61(2):69e90.
[2] Mayer EL, Burstein HJ. Postmenopausal breast cancer: a best
endocrine strategy? Lancet 2015;386(10001):1317e9.
[3] Early Breast Cancer Trialists’ Collaborative G, Dowsett M,
Forbes JF, Bradley R, Ingle J, Aihara T, et al. Aromatase in-
hibitors versus tamoxifen in early breast cancer: patient-level
meta-analysis of the randomised trials. Lancet 2015;386(10001):
1341e52.
[4] Rossi E, Morabito A, Di Rella F, Esposito G, Gravina A,
Labonia V, et al. Endocrine effects of adjuvant letrozole
compared with tamoxifen in hormone-responsive postmenopausal
patients with early breast cancer: the HOBOE trial. J Clin Oncol
2009;27(19):3192e7.
[5] Gonnelli S, Petrioli R. Aromatase inhibitors, efficacy and meta-
bolic risk in the treatment of postmenopausal women with early
breast cancer. Clin Interv Aging 2008;3(4):647e57.
[6] Younossi ZM, Koenig AB, Abdelatif D, Fazel Y, Henry L,
Wymer M. Global epidemiology of nonalcoholic fatty liver
disease-meta-analytic assessment of prevalence, incidence, and
outcomes. Hepatology 2016;64(1):73e84.
114 N. Hong et al. / European Journal of Cancer 82 (2017) 103e114
[7] Byrne CD, Targher G. NAFLD: a multisystem disease. J Hepatol
2015;62(1 Suppl.):S47e64.
[8] Farrell GC. Drugs and steatohepatitis. Semin Liver Dis 2002;
22(2):185e94.
[9] Fromenty B. Drug-induced liver injury in obesity. J Hepatol 2013;
58(4):824e6.
[10] Liu CL, Huang JK, Cheng SP, Chang YC, Lee JJ, Liu TP. Fatty
liver and transaminase changes with adjuvant tamoxifen therapy.
Anticancer Drugs 2006;17(6):709e13.
[11] Lin Y, Liu J, Zhang X, Li L, Hu R, Liu J, et al. A prospective,
randomized study on hepatotoxicity of anastrozole compared
with tamoxifen in women with breast cancer. Cancer Sci 2014;
105(9):1182e8.
[12] Katznelson L, Riskind PN, Saxe VC, Klibanski A. Prolactin
pulsatile characteristics in postmenopausal women. J Clin Endo-
crinol Metab 1998;83(3):761e4.
[13] Ballestri S, Romagnoli D, Nascimbeni F, Francica G, Lonardo A.
Role of ultrasound in the diagnosis and treatment of nonalcoholic
fatty liver disease and its complications. Expert Rev Gastroenterol
Hepatol 2015;9(5):603e27.
[14] Lee JH, Kim D, Kim HJ, Lee CH, Yang JI, Kim W, et al. Hepatic
steatosis index: a simple screening tool reflecting nonalcoholic
fatty liver disease. Dig Liver Dis 2010;42(7):503e8.
[15] Matthews DR, Hosker JP, Rudenski AS, Naylor BA,
Treacher DF, Turner RC. Homeostasis model assessment: insulin
resistance and beta-cell function from fasting plasma glucose and
insulin concentrations in man. Diabetologia 1985;28(7):412e9.
[16] Wai CT, Greenson JK, Fontana RJ, Kalbfleisch JD, Marrero JA,
Conjeevaram HS, et al. A simple noninvasive index can predict
both significant fibrosis and cirrhosis in patients with chronic
hepatitis C. Hepatology 2003;38(2):518e26.
[17] Chalasani N, Younossi Z, Lavine JE, Diehl AM, Brunt EM,
Cusi K, et al. The diagnosis and management of non-alcoholic
fatty liver disease: practice guideline by the American Associa-
tion for the Study of Liver Diseases, American College of
Gastroenterology, and the American Gastroenterological Asso-
ciation. Hepatology 2012;55(6):2005e23.
[18] European Association for the Study of the Liver. Electronic
address eee, European Association for the Study of D, European
Association for the Study of O. EASL-EASD-EASO clinical
practice guidelines for the management of non-alcoholic fatty
liver disease. J Hepatol 2016;64(6):1388e402.
[19] King PD, Perry MC. Hepatotoxicity of chemotherapy. Oncologist
2001;6(2):162e76.
[20] Ali MS, Groenwold RH, Belitser SV, Pestman WR, Hoes AW,
Roes KC, et al. Reporting of covariate selection and balance
assessment in propensity score analysis is suboptimal: a systematic
review. J Clin Epidemiol 2015;68(2):112e21.
[21] Stuart EA. Matching methods for causal inference: a review and a
look forward. Stat Sci 2010;25(1):1e21.
[22] Murata Y, Ogawa Y, Saibara T, Nishioka A, Fujiwara Y,
Fukumoto M, et al. Unrecognized hepatic steatosis and non-
alcoholic steatohepatitis in adjuvant tamoxifen for breast cancer
patients. Oncol Rep 2000;7(6):1299e304.
[23] Ferrara CM, Lynch NA, Nicklas BJ, Ryan AS, Berman DM.
Differences in adipose tissue metabolism between postmenopausal
and perimenopausal women. J Clin Endocrinol Metab 2002;87(9):
4166e70.
[24] Herrington DM, Klein KP. Effects of SERMs on important in-
dicators of cardiovascular health: lipoproteins, hemostatic
factors, and endothelial function. Womens Health Issues 2001;
11(2):95e102.
[25] Perez EA. Safety profiles of tamoxifen and the aromatase in-
hibitors in adjuvant therapy of hormone-responsive early breast
cancer. Ann Oncol 2007;18(Suppl. 8). viii26e35.
[26] Francque SM, van der Graaff D, Kwanten WJ. Non-alcoholic
fatty liver disease and cardiovascular risk: pathophysiological
mechanisms and implications. J Hepatol 2016;65(2):425e43.
[27] Ekstedt M, Hagstrom H, Nasr P, Fredrikson M, Stal P,
Kechagias S, et al. Fibrosis stage is the strongest predictor for
disease-specific mortality in NAFLD after up to 33 years of
follow-up. Hepatology 2015;61(5):1547e54.
[28] Kim D, Kim WR, Kim HJ, Therneau TM. Association between
noninvasive fibrosis markers and mortality among adults with
nonalcoholic fatty liver disease in the United States. Hepatology
2013;57(4):1357e65.
[29] McPherson S, Hardy T, Henderson E, Burt AD, Day CP,
Anstee QM. Evidence of NAFLD progression from steatosis to
fibrosing-steatohepatitis using paired biopsies: implications for
prognosis and clinical management. J Hepatol 2015;62(5):
1148e55.
[30] Singh S, Allen AM, Wang Z, Prokop LJ, Murad MH, Loomba R.
Fibrosis progression in nonalcoholic fatty liver vs nonalcoholic
steatohepatitis: a systematic review and meta-analysis of paired-
biopsy studies. Clin Gastroenterol Hepatol 2015;13(4):643e54.
e1-9; quiz e39-40.
[31] Nguyen MC, Stewart RB, Banerji MA, Gordon DH, Kral JG.
Relationships between tamoxifen use, liver fat and body fat dis-
tribution in women with breast cancer. Int J Obes Relat Metab
Disord 2001;25(2):296e8.
[32] Liu CL, Yang TL. Sequential changes in serum triglyceride levels
during adjuvant tamoxifen therapy in breast cancer patients and
the effect of dose reduction. Breast Cancer Res Treat 2003;79(1):
11e6.
[33] Gudbrandsen OA, Rost TH, Berge RK. Causes and prevention of
tamoxifen-induced accumulation of triacylglycerol in rat liver. J
Lipid Res 2006;47(10):2223e32.
[34] Begriche K, Massart J, Robin MA, Borgne-Sanchez A,
Fromenty B. Drug-induced toxicity on mitochondria and lipid
metabolism: mechanistic diversity and deleterious consequences
for the liver. J Hepatol 2011;54(4):773e94.
[35] Rabinowich L, Shibolet O. Drug induced steatohepatitis: an un-
common culprit of a common disease. Biomed Res Int 2015;2015:
168905.
[36] Kwon HK, Greenson JK, Conjeevaram HS. Effect of lifetime
alcohol consumption on the histological severity of non-alcoholic
fatty liver disease. Liver Int 2014;34(1):129e35.
[37] Dunn W, Sanyal AJ, Brunt EM, Unalp-Arida A, Donohue M,
McCullough AJ, et al. Modest alcohol consumption is associated
with decreased prevalence of steatohepatitis in patients with non-
alcoholic fatty liver disease (NAFLD). J Hepatol 2012;57(2):
384e91.
[38] Hershman DL, Kushi LH, Shao T, Buono D, Kershenbaum A,
Tsai WY, et al. Early discontinuation and nonadherence to
adjuvant hormonal therapy in a cohort of 8,769 early-stage breast
cancer patients. J Clin Oncol 2010;28(27):4120e8.
[39] Mishra P, Younossi ZM. Abdominal ultrasound for diagnosis of
nonalcoholic fatty liver disease (NAFLD). Am J Gastroenterol
2007;102(12):2716e7.