ORIGINAL ARTICLE—LIVER, PANCREAS, AND BILIARY TRACT
Oral contraceptive pill use is associated with reduced odds
of nonalcoholic fatty liver disease in menstruating women: results
from NHANES III
Su-Hsun Liu•Mariana Lazo•Ayman Koteish•
W. H. Linda Kao•Ming-Hsiung Shih•Susanne Bonekamp•
Ruben Hernaez•Jeanne M. Clark
Received: 7 August 2012/Accepted: 5 November 2012/Published online: 28 November 2012
? Springer Japan 2012
liver disease (NAFLD) in men and postmenopausal women
than in premenopausal women has suggested a potential
role of sex hormones in the pathogenesis of the disease. We
sought to evaluate the association between oral contra-
ceptive pills (OCP) and NAFLD and to determine whether
adiposity mediates any effect.
MethodsWe included 4338 women aged 20–60 years
who were enrolled in the Third National Health and
Nutrition Examination Survey from 1988 to 1994 in a
population-based cross-sectional study. We defined NA-
FLD as moderate–severe steatosis on ultrasonography in
women without excessive alcohol use or other identifiable
Higher prevalence of nonalcoholic fatty
causes. OCP use was based on self-report and was cate-
gorized as never, former or current use.
ResultsThe overall weighted prevalence of NAFLD was
11.6 % but lower in current (6.7 %) than in former
(12.0 %) or never users (15.6 %, P = 0.016). In the mul-
tivariable model, current OCP users experienced a 50 %
lower odds of NAFLD than never users (adjusted odds ratio
0.50; 95 % confidence interval 0.26, 0.98) after adjusting
for age, race/ethnicity, smoking status, history of diabetes
or hypertension and education. Further adjustment for body
mass index or waist circumference significantly attenuated
the OCP–NAFLD relationship.
ConclusionsIn this large US-representative population,
OCP use was associated with reduced odds of NAFLD.
However, this association could be mediated or con-
founded by adiposity. Prospective studies are needed to
further clarify the causal role of sex hormone.
Obesity ? Oral contraceptive pill ? Sex hormone
Adiposity ? Nonalcoholic fatty liver disease ?
Nonalcoholic fatty liver disease (NAFLD) has become the
most common liver disease in developed countries over the
past two decades. Due to an increased prevalence of
obesity, type 2 diabetes and hyperlipidemia, NAFLD and
its liver-related consequences are expected to become
public health problems, and thus identification of factors
that can prevent the development of NAFLD is key [1, 2].
While higher prevalence of NAFLD is commonly found
in Hispanics than in non-Hispanic whites  or in African
Americans [3–5], NAFLD has largely been associated
with older age, inflammatory biomarkers involved in
S.-H. Liu ? M. Lazo ? W. H. L. Kao ? R. Hernaez ?
J. M. Clark (&)
Department of Epidemiology, Johns Hopkins Bloomberg
School of Public Health, 2024 E. Monument St.,
Suite 2-600, Baltimore, MD 21287, USA
A. Koteish ? R. Hernaez ? J. M. Clark
Department of Medicine, Johns Hopkins University
School of Medicine, Baltimore, MD, USA
Department of Family Medicine, Far Eastern Memorial Hospital,
New Taipei City, Taiwan
Russell H. Morgan Department of Radiology and Radiological
Science, Johns Hopkins University School of Medicine,
Baltimore, MD, USA
Department of Medicine, Washington Hospital Center/
Georgetown University Hospital, Washington, DC, USA
J Gastroenterol (2013) 48:1151–1159
hepatocellular injury, impaired glucose and lipoprotein
metabolism and anthropometric measures that reflect cen-
tral obesity [1, 2]. In addition, NAFLD is most frequently
found in men and postmenopausal women without hor-
mone replacement therapy (HRT) , suggesting a role of
sex hormones, particularly estrogen, in the pathogenesis of
NAFLD [6, 7].
Animal studies and clinical observations indicate that
high levels of estrogen, as normally seen in premenopausal
women, preferentially increase peripheral distribution of
fat, thereby reducing the risk of central obesity; whereas the
fall in the levels of estrogen observed during menopause, is
associated with loss of subcutaneous fat and a gain in vis-
ceral fat [8–10]. Similarly, observational studies have
described more pronounced deposition of fat mass in the
waist relative to the hip and high waist-to-hip ratio in
postmenopausal compared to premenopausal women [11,
12]. Previous studies have reported that menopause
increased the risk of NAFLD and that postmenopausal
women on HRT were less likely to have NAFLD than those
not treated with hormones [9, 13–15]. However, there is
limited evidence on whether the use of oral contraceptive
pills (OCP), another source of exogenous hormones, can
also protect younger women from the risk of NAFLD.
Therefore, we sought to determine the association between
OCP use and prevalent NAFLD in women aged 20–60 who
were enrolled in the Third National Health and Nutrition
Examination Survey (NHANES III) from 1988 to 1994. We
hypothesized that women taking OCPs would have lower
odds of NAFLD than those who never used OCP. We also
examined whether the association could be explained by the
postulated effects of hormones on adiposity.
Study design and population
NHANES III was conducted between 1988 and 1994 as
part of a program consisting of cross-sectional population-
based studies to assess the health and nutritional status of
non-institutionalized adults and children in the US.
Extensive documentation regarding the data collection
procedure and contents of questionnaire for adults in
NHANES III is available elsewhere . Briefly, in addi-
tion to demographic, socioeconomic and dietary informa-
tion, detailed lifestyle and behavioral questionnaire for
adults aged 20 or older were documented through house-
hold interview as well as comprehensive physical exam-
inations and blood samples collection performed in mobile
examination centers (MEC). Overall, 33,944 participants
were interviewed in NHANES III, of which 14,797 adults
aged 20–74 completed household interview, physical
examinations and a gallbladder ultrasonography. For the
current study, we included 4338 females aged 20–60 whose
anthropometric measures were available and who provided
their reproductive health information regarding the last
menstrual period (LMP) and OCP use. Women who had
LMP within the prior 12 months were considered regularly
menstruating (Fig. 1). Considering the distinct profiles of
endogenous hormones in women of twenties through six-
ties, we further grouped women into a younger, reproduc-
tive group (age 20–34) and a mid-adult, perimenopausal
group (age 35–60). Detailed menstruation history was
unavailable for staging women’s perimenopausal status
according to the 2011 Stages of Reproductive Aging
Workshop (STRAW) ? 10 criteria .
In 2009–2010, hepatic steatosis was systematically deter-
mined from archived gallbladder ultrasounds using modi-
fied criteria by Hamaguchi et al.  and included the
assessment of: liver parenchyma brightness, liver-to-kid-
ney contrast, deep beam attenuation, bright vessel walls
and thicknessof gallbladder
descriptions about the review protocol are available else-
where . NAFLD was defined as moderate or severe
hepatic steatosis in women who did not use zidovudine or
didanosine and who reported daily consumption of one
alcohol drink (14 g alcohol) or less in the prior 12 months
. The intra- and inter-rater reliability of the ultrasound-
defined hepatic steatosis was 0.77 [95 % confidence
interval (CI) 0.73–0.82] and 0.70 (95 % CI 0.64–0.76),
Women aged 20–60 were asked: ‘‘Have you ever taken
birth-control/estrogen pills?’’, ‘‘How old when you first
took birth-control/estrogen pills?’’, ‘‘How many months
ago did you stop taking birth control pills or are you still
taking them?’’ and ‘‘Not counting any time when you
stopped taking them, for how many months have you
taken/did you take birth-control/estrogen pills?’’ Based on
responses to these answers, women were categorized as:
never, former or current users of OCP. In addition, we
examined the age of initiation, the duration of OCP use
(excluding intervals not taking the pills among former and
current users), and time since stopping OCP (in months) in
former users only.
Descriptive analyses were performed to compare socio-
demographic characteristics and laboratory profiles among
participants who never, formerly or currently used OCP at
the time of the household interview. To assess unadjusted
associations, design-based chi-square statistics were used
1152 J Gastroenterol (2013) 48:1151–1159
for categorical variables, and survey-based linear regres-
sion analyses were conducted for continuous covariates.
Covariates that were considered in multivariable models
included age, race/ethnicity (non-Hispanic white, non-
Hispanic black, Mexican-American and others), past
medical history of diabetes mellitus (DM) or hypertension,
smoking behavior (never, former or current smokers),
socio-economic indicators such as highest educational
level obtained (high school or higher vs. less than high
school), insurance coverage status (yes vs. no) and income
level (at or above poverty ratio, 1 vs. below poverty ratio,
1); liver enzymes including alanine aminotransferase
(ALT) and aspartate aminotransferase (AST) as well as
plasma levels of glucose and insulin. Among women who
fasted for 8 h or longer, we also compared serum levels of
triglyceride (TG) by OCP use and estimated the degree of
insulin resistance by calculating the homeostasis model
assessment-estimated insulin resistance index (HOMA-IR)
. Stratified analysis and formal statistical tests for
interaction were used to decide whether OCP–NAFLD
association differed by age, race, DM or hypertension
history. When no such effect modification was identified
among the above mentioned variables, multivariable
logistic regression was then built to determine the inde-
pendent association between OCP use and NAFLD while
adjusting for all potential confounders.
In the mediation analysis, anthropometric variables that
measure different aspects of adiposity, such as body mass
index (BMI), waist circumference (WC), waist-to-hip ratio
and total body percent fat , were first examined by
inclusion of each variable into the base multivariable
model to assess whether the measure of association
between OCP use and NAFLD was attenuated by at least
10–15 % on the log scale . Assumptions for valid
mediation analysis, such as whether OCP users were dis-
tributed equally across different levels of each of the
potential mediators, were examined empirically [23, 24]
before calculating the absolute proportion of exposure
effect explained (aPEE) for each potential mediator. Odds
ratios (OR) and 95 % CI were reported at a two-tailed
significance level of 0.05. All analyses were performed in
Stata 11 , employing appropriate sampling weights to
account for the stratified, probability sampling and non-
responses in the survey.
Characteristics of subjects
Characteristics of the study population and by OCP use
history are summarized in Table 1. Compared to never
users, current OCP users were younger (mean 27.6 vs.
34.1 years), less likely to be Mexican-American (5.3 vs.
9.5 %) or to have diabetes (2.0 vs. 5.6 %) or hypertension
(11.6 vs. 17.9 %). Current users were also more likely to
have a higher educational attainment (88.3 vs. 75.9 %) and
income (86.6 vs. 79.6 % were at or above poverty line);
Fig. 1 Flowchart of participant
selection. Among 33,944
participants who were
interviewed in NHANES III in
1988–1994, 14,797 adults aged
20–74 completed adult
questionnaire (AQ) and
received physical examinations
at the mobile examination
center (MEC) where blood
samples collection and a
were also performed. In the
current analysis, we included
4338 females aged 20–60 who
reported their history of oral
contraceptive pills use and had
the last menstrual period (LMP)
within the prior 12 months.
Women without any of the
measures were excluded: body
mass index, waist circumference
or waist-to-hip ratio
J Gastroenterol (2013) 48:1151–11591153
Table 1 Characteristics of premenopausal women aged 20–60 in NHANES III (N = 4338) by OCP Use History, 1988–1994
All (N = 4338) Never use
(N = 960)
(N = 2604)
(N = 774)
NAFLD11.6 % 15.6 %12.0 % 6.7 %
Age (years): mean (SEa) 33.7 (0.3)34.1 (0.6)35.5 (0.2)27.6 (0.4)
20–34 55.0 % 54.2 %45.0 %87.7 %
35–60 45.1 %45.9 %55.0 % 12.3 %
Non-Hispanic white72.9 % 59.3 %75.7 %76.8 %
Non-Hispanic black 12.3 %11.7 %12.8 % 11.6 %
Mexican-American5.7 %9.5 % 4.8 % 5.3 %
Others9.0 %19.5 %6.7 %6.3 %
High school or higher83.4 % 75.9 %84.2 % 88.3 %
Income leveln = 3996n = 849n = 2431n = 716
At or above poverty line85.2 %79.6 % 86.4 %86.6 %
Insurance coveragen = 4148n = 913n = 2491n = 744
Yes 86.5 %80.9 %87.5 %88.5 %
25.8 (0.2)26.6 (0.7) 26.0 (0.2) 24.1 (0.3)
55.8 % 52.9 %52.9 %67.9 %
22.5 %21.0 % 24.7 % 17.2 %
C30 21.7 %26.1 % 22.4 %14.9 %
Waist circumference (cm)
35.2 % 40.8 %38.1 %20.6 %
C0.8 69.8 %72.4 % 74.2 %53.4 %
Total percent body fat (%)n = 3808n = 831n = 2255n = 722
C4022.3 24.624.3 14.1
Risk factors (%)
Diabetes 4.4 5.64.7 2.0
Hypertension18.6 17.921.0 11.6
Never 55.866.1 51.659.3
Former smoker 16.19.7 18.315.4
Current smoker 28.1 24.330.1 25.3
Cholesterol (mg/dL)n = 4133n = 909n = 2487n = 737
9.9 %10.1 % 9.9 %9.6 %
n = 2674n = 594n = 1593n = 487
C150 mg/dL16.0 %18.3 %15.9 % 14.0 %
n = 4102n = 900n = 2472n = 730
3.4 %3.9 %3.7 %2.0 %
n = 4102n = 900n = 2472n = 730
4.4 %6.6 %4.5 % 2.2 %
n = 4167 n = 915n = 2512n = 740
Serum insulin (lU/mL)
0.6 % 0.7 % 0.7 % 0.1 %
n = 4136n = 912n = 2491n = 733
9.6 (0.2) 11.0 (0.7)9.4 (0.2)9.1 (0.3)
1154 J Gastroenterol (2013) 48:1151–1159
access to care as indicated by insurance coverage (88.5 vs.
80.9 %); as well as being current or former smokers (40.7
vs. 34.0 %). In addition, more women taking OCPs had a
BMI \25 kg/m2(67.9 vs. 52.9 %) or a normal level of
serum ALT (B31 U/L) than those who never used OCPs
(97.8 vs. 93.4 %). However, current OCP users were less
likely to have an elevated serum concentration of insulin or
insulin resistance (HOMA-IR C2.60) than never users
(14.4 vs. 21.6 %) whereas former users had a comparable
level of HOMA-IR to never users. Women had similar
levels of serum TG and plasma glucose regardless of OCP
use status. As shown in Table 2, current OCP users began
to take OCP at a similar age to those who reported former
use (20.0 vs. 20.2 years). However, 19.9 % of current users
reported having been on OCP for more than a decade while
21.7 % of former users had stopped taking oral pills within
the first year of use (Table 2).
Associations between OCP and NAFLD
Overall, a protective effect of current OCP use remained
significant after adjusting for age, race/ethnicity, smoking
history, co-morbidities and the highest educational level
(Model 1 in Table 3), with a 50 % lower odds of NAFLD
in current versus never users [adjusted odds ratio (AOR)
0.50; 95 % CI 0.26, 0.98]. In Model 2, obesity as defined
by BMI C30 kg/m2was found to be strongly associated
with NAFLD compared to their normal-BMI counterparts,
overweight females showed a 1.3-fold odds of NAFLD
(AOR 1.31; 95 % CI 0.91, 1.89) and obese individuals had
a 3.3-fold higher odds of NAFLD (AOR 4.27; 95 % CI
2.88, 6.34). BMI significantly attenuated the relationship
between current OCP use and NAFLD (AOR 0.55; 95 %
CI 0.27, 1.13), by about 14 % [absolute PEE (aPEE) was
14.0 %]. In other words, slightly more than one-tenth of the
association between OCP use and NAFLD could have been
explained by BMI alone. Similarly, women with a WC
[88 cm (34.6 in.) also exhibited a twofold higher odds of
NAFLD than those with smaller waist (AOR 3.21; 95 % CI
2.25, 4.58). WC also substantially attenuated the OCP–
NAFLD association with an estimated aPEE 21.3 %
(Model 3 in Table 3). On the other hand, the waist-to-hip
ratio and total percent body fat appeared to exert only
minimal influence (aPEE \15 %) (Models 4 and 5,
The low prevalence of NAFLD associated with current
OCP use seemed particularly prominent in women aged
20–34 (AOR 0.41, 95 % CI 0.20, 0.86) but not in women
aged 35 or older (AOR 1.58, 95 % CI 0.51, 4.91).
Table 2 Former and current OCP users among premenopausal
women, NHANES III, 1988–1994
OCP use Former use
(N = 2604)
(N = 774)
Age (years): mean (SEa) 35.5 (0.2) 27.6 (0.4)
Age at first use of OCP (years): mean
Years of taking OCP: mean (SEa)
20.2 (0.1) 20.0 (0.3)
4.3 (0.2) 6.7 (0.6)
21.7 %8.0 %
41.0 %37.7 %
5–1024.7 %34.4 %
Years since stopping OCP: mean (SEa)
12.6 % 19.9 %
5–10 20.8 %
OCP oral contraceptive pill, SD standard deviation
aSurvey-based mean estimation with linearized standard error
Table 1 continued
All (N = 4338)Never use
(N = 960)
(N = 2604)
(N = 774)
69.7 % 68.0 %70.1 %70.2 %
24.4 %22.9 %24.3 % 26.3 %
5.9 %9.2 % 5.7 % 3.5 %
n = 2528n = 540n = 1463n = 451
HOMA-IR C2.620.6 %21.6 %22.3 %14.4 %
ALT alanine aminotransferase, AST aspartate aminotransferase, BMI body mass index, HOMA-IR the homeostasis model assessment-estimated
insulin resistance index, OCP oral contraceptive pill, SE standard error
aSurvey-based mean estimation with linearized standard error
bOnly included participants who fasted C8 h
c3.9 % of participants missing plasma level of glucose; included participants who fasted C8 h (65.1 %),\8 h (34.8 %) and 4 participants with
unknown fasting status
dCalculated only among participants who fasted C8 h and had plasma levels of glucose and insulin available (58.3 %)
J Gastroenterol (2013) 48:1151–11591155
However, there was no statistical evidence for the differ-
ential effect of age on the OCP–NAFLD association
(P values for interaction[0.05). Nor did the OCP–NAFLD
association significantly differ among women with or
without diabetes or hypertension (P values for interaction
[0.05, data not shown). Moreover, we found that the
association between OCP use and NAFLD was differen-
tially affected by race, with a protective effect of current
OCP use observed only among non-Hispanic White (AOR
0.41, 95 % CI 0.18, 0.97) after adjusting for age, smoking
history, co-morbidities and the highest educational level
(Table 4). Notably, African American women who never
Table 3 Odds ratios of NAFLD by oral contraceptive pill use, premenopausal women in NHANES III (N = 4338), 1988–1994
Adjusted for age,
AOR (95 % CI)
Adjusted for age,
AOR (95 % CI)
Adjusted for age,
AOR (95 % CI)
Adjusted for age,
AOR (95 % CI)
Adjusted for age,
AOR (95 % CI)
Former 0.81 (0.54, 1.20) 0.84 (0.54, 1.28)0.83 (0.55, 1.26)0.80 (0.53, 1.19) 0.86 (0.57, 1.30)
Current 0.50 (0.26, 0.98)0.55 (0.27, 1.13) 0.58 (0.29, 1.17) 0.52 (0.26, 1.03)0.53 (0.24, 1.16)
0.0310.0870.104 0.042 0.095
1.31 (0.91, 1.89)
4.27 (2.88, 6.34)
3.21 (2.25, 4.58)
1.40 (1.03, 1.91)
PBF C40 %
2.67 (1.88, 3.78)
– 14.0 %21.3 % 5.4 % 12.4 %
AOR adjusted odds ratio, aPEE absolute proportion of exposure effect, BMI body mass index, CI confidence interval, DM self-reported diagnosis
of type 2 diabetes, PBF total percent body fat, WC waist circumference, WHR waist-to-hip ratio
aWomen without bio-impedance measure were not included (n = 530)
bAge was entered as a continuous variable; race grouped as non-Hispanic white, non-Hispanic black, Mexican-American and others; smoking as
a categorical variable that indicated never, former and current smoking status; education as a binary indicator for a high-school or higher
cAbsolute proportion of exposure effect (aPEE) was estimated by calculating percent change in regression coefficients of current versus never
users between the extended and the base models, using the same estimation sample
Table 4 OCP–NAFLD association by race, premenopausal women aged 20–60 years in NHANES III (N = 4338), 1988–1994
Non-Hispanic whiteNon-Hispanic blackMexican-AmericanOthers
95 % CIAORa
95 % CI AORa
95 % CIAORa
95 % CI
Never1–0.46 0.25, 0.84b
1.60 0.96, 2.691.30 0.60, 2.81
Former* 0.740.41, 1.33
0.770.43, 1.40 1.300.75, 2.24 0.870.31, 2.47
0.680.26, 1.82 1.01 0.55, 1.85 0.610.15, 2.44
AOR adjusted odds ratio, CI confidence interval, OCP oral contraceptive pill
* P values for interaction: 0.024 for former versus never users; 0.011 for current versus never users
aAdjusted for age, smoking status, history of diabetes or hypertension and highest educational level
bP value for interaction 0.04
cP value for interaction 0.017
1156J Gastroenterol (2013) 48:1151–1159
used OCP had a 54 % lower odds of NAFLD than their
non-Hispanic White counterparts (AOR 0.46, 95 % CI
In this population-based study, we found an independent
association between current OCP use and NAFLD among
women of reproductive age. This study finding extends the
accumulating literature that exogenous sex hormones may
provide a protective effect against NAFLD in women.
Current literature on the relationship of exogenous sex
hormones and adiposity remains inconclusive [2, 6, 12, 15].
Particularly, in a recent systematic review, most included
trials were of short duration (\1 year)  whereas more
than 90 % of current and 75 % of former OCP users in the
current study reported that they had taken OCP for more
than 1 year at the time of the interview or before quitting.
As such, we hypothesized that adiposity situated on the
pathway between OCP use and NAFLD as a mediator,
rather than as a confounder. Nevertheless, the directionality
of the OCP-obesity relationship can only be clarified in a
While female sex hormones are known to modulate
immune responses underlying some viral infections and
chronic inflammatory disorders [27, 28], several mecha-
nisms have been proposed by which estrogen may exert a
hepatoprotective effect [7, 27]. Both in vitro and clinical
studies have shown that estradiol (E2) at the pregnancy
level can block the secretion of proinflammatory cytokines
such as tumor necrosis factor (TNF)-a and interleukin
(IL)-6 [27, 29, 30]. E2 is also known to be potent anti-
oxidant acting through estrogen receptors on hepatic cells
. Conceivably, low levels of E2 may be associated
with impaired beta-oxidation of fatty acids that are
deposited in hepatocytes, worsening the steatosis in
patients with NAFLD [7, 32]. In contrast, research on the
development and differentiation has been relatively little
and the results are often inconsistent [33, 34]. Future
investigations to further delineate the relative contribu-
tions of estrogen and progesterone in the pathogenesis of
NAFLD are warranted.
Given the low prevalence of current OCP users among
participantsaged35orolder(weightedprevalence5.3 % vs.
30.8 % in those aged 20–34 years), our finding that the
protective effect of current OCP use against NAFLD was
most evident in younger (age \35 years) versus older
women was not unexpected. This also likely explains our
lack of power to detect a significant modifying effect of age
on the OCP–NAFLD association. As early work has
revealed that women taking OCPs generally have a low
concentration of plasma estrogen [35, 36], we believe that
endogenous estrogen may not be involved in this differ-
ential relationship by age in current OCP users.
Consistent with other work showing that African
Americans tended to have a reduced risk of nonalcoholic
steatohepatitis (NASH) [3, 37], we found that African
American women who never used OCP had a reduced odds
of NAFLD as compared to non-Hispanic White. Possible
explanations for this risk reduction by race abound yet
none is more favorable than others. Among a cohort of
obese women requiring bariatric surgery, Solga et al. 
found that African American women were less likely to be
affected by moderate–severe hepatic steatosis or liver
fibrosis than their Caucasian counterparts, despite of sim-
ilarly high prevalence of risk factors for cardio-metabolic
disorders. Solga et al.  suggested a gene–environmental
interaction that modified women’s immune responses to an
altered adiposity status. Besides, several observational
studies have demonstrated that African American women
tend to have less visceral fat than Caucasian women of
similar BMI values, suggesting race-specific metabolic
pathways in utilization and storage of adipose tissue [39,
40]. As studies have shown that E2 levels were generally
comparable among African American women, non-His-
panic White and Hispanic women [41–43], it is possible
that estrogen differentially modulates the immune system
under an oxidative stress in the liver of African American
women as compared to non-Hispanic Whites. However,
whether this OCP–NAFLD association truly differs by race
needs further research. Longitudinal studies that monitor
changes in immune cells and cytokines along with hor-
mone levels in a multi-racial population may help elucidate
the interplay among the immune, endocrine and metabolic
aspects of this particular liver disease.
Several limitations need to be considered in light of our
findings. First, the cross-sectional nature makes it impos-
sible to ascertain causality. Nevertheless, our findings
contribute to the existing literature that exogenous sex
hormone may independently protect women against NA-
FLD by providing evidence from a large, representative
female population in the US. Second, while performing
mediation analysis, theoretical work in the causal inference
has shown that it is imperative to make sure the mediator-
outcome relationship is well-adjusted by including most, if
not all, measured confounders in the statistical model .
Although NHANES III has extremely comprehensive data
collection, the risk for omitted variables still exists and
cannot be determined by empirical analysis. Lastly,
women’s exposure to OCP relied on self-report in the
current analysis. Regiments or doses of OCP were not
available to verify the relative contribution of estrogen
versus progesterone or to evaluate the dose– or duration–
response relationship with NAFLD.
J Gastroenterol (2013) 48:1151–11591157
In this US-representative sample of women aged
20–60 years, we found that current use of OCP was inde-
pendently associated with decreased odds of NAFLD. A
significant proportion of such effect could be attributed to
adiposity as measured by waist measure or BMI. Addi-
tional investigations using prospective cohorts, with very
well characterized measures of adiposity are needed to
determine whether a causal relationship between OCP use
and NAFLD exists.
Conflict of interest
The authors declare that they have no conflict
1. Angulo P. Nonalcoholic fatty liver disease. NEJM. 2002;346:
2. Lazo M, Clark JM. The epidemiology of nonalcoholic fatty liver
disease: a global perspective. Semin Liver Dis. 2008;28:339–50.
3. Browning JD, Szczepaniak LS, Dobbins R, Nuremberg P, Horton
JD, Cohen JC, et al. Prevalence of hepatic steatosis in an urban
population in the United States: impact of ethnicity. Hepatology.
4. Clark JM. The epidemiology of nonalcoholic fatty liver disease in
adults. J Clin Gastroenterol. 2006;40(Suppl 1):S5–10.
5. Mohanty SR, Troy TN, Huo D, O’Brien BL, Jensen DM, Hart J.
Influence of ethnicity on histological differences in non-alcoholic
fatty liver disease. J Hepatol. 2009;50:797–804.
6. Gutierrez-Grobe Y, Ponciano-Rodriguez G, Ramos MH, Uribe
M, Mendez-Sanchez N. Prevalence of non alcoholic fatty liver
disease in premenopausal, postmenopausal and polycystic ovary
syndrome women. The role of estrogens. Ann Hepatol. 2010;9:
7. Shimizu I, Ito S. Protection of estrogens against the progression
of chronic liver disease. Hepatol Res. 2007;37:239–47.
8. Elbers JM, de Roo GW, Popp-Snijders C, Nicolaas-Merkus A,
Westerveen E, Joenje BW, et al. Effects of administration of
17beta-oestradiol on serum leptin levels in healthy postmeno-
pausal women. Clin Endocrinol (Oxf). 1999;51:449–54.
9. Haarbo J, Marslew U, Gotfredsen A, Christiansen C. Postmeno-
pausal hormone replacement therapy prevents central distribution
of body fat after menopause. Metabolism. 1991;40:1323–6.
10. Witteman J, Grobbee D, Kok F, Hofman A, Valkenburg H.
Increased risk of atherosclerosis in women after the menopause.
Br Med J. 1989;298:642–4.
11. Bjoerkelund C, Lissner L, Andersson S, Lapidus L, Bengtsson C.
Reproductive history in relation to relative weight and fat dis-
tribution. Int J Obes. 1996;20:213–9.
12. Saruc ¸ M, Yu ¨ceyar H, Ayhan S, Tu ¨rkel N, Tuzcuoglu I, Can M.
The association of dehydroepiandrosterone, obesity, waist–hip
ratio and insulin resistance with fatty liver in postmenopausal
women—a hyperinsulinemic euglycemic insulin clamp study.
13. Clark JM, Brancati FL, Diehl AM. Nonalcoholic fatty liver dis-
ease. Gastroenterology. 2002;122:1649–57.
14. McKenzie J, Fisher BM, Jaap AJ, Stanley A, Paterson K, Sattar
N. Effects of HRT on liver enzyme levels in women with type 2
diabetes: a randomized placebo-controlled trial. Clin Endocrinol
15. Park SH, Jeon WK, Kim SH, Kim HJ, Park DI, Cho YK, et al.
Prevalence and risk factors of non-alcoholic fatty liver disease
among Korean adults. J Gastroenterol Hepatol. 2006;21:138–43.
16. National Center for Health Statistics. NHANES III data files,
documentation and codebooks; 2011. http://www.cdc.gov/nchs/
nhanes/nh3data.htm. Accessed 16 Nov 2012.
17. Harlow SD, Gass M, Hall JE, Lobo R, Maki P, Rebar RW, et al.
Executive summary of the Stages of Reproductive Aging
Workshop ? 10: addressing the unfinished agenda of staging
reproductive aging. J Clin Endocrinol Metab. 2012;97:1159–68.
18. Hamaguchi M, Kojima T, Itoh Y, et al. The severity of ultraso-
nographic findings in nonalcoholic fatty liver disease reflects the
metabolic syndrome and visceral fat accumulation. Am J Gas-
19. Centers for Disease Control and Prevention. Alcohol and Public
Health. Frequently asked questions: what is a standard drink in
the United States? 2012. http://www.cdc.gov/alcohol/faqs.htm#
standDrink . Accessed 16 Nov 2012.
20. National Center for Health Statistics. Third National Health and
Nutrition Survey: hepatic steatosis assessment procedure manual.
Steatosis_Ultrasound_Procedures_Manual.pdf. Accessed on 16
21. Bonora E, Targher G, Alberiche M, Bonadonna RC, Saggiani F,
Zenere MB, et al. Homeostasis model assessment closely mirrors
the glucose clamp technique in the assessment of insulin sensi-
tivity: studies in subjects with various degrees of glucose toler-
ance and insulin sensitivity. Diabetes Care. 2000;23:57–63.
22. Chumlea WC, Guo SS, Kuczmarski RJ, Flegal KM, Johnson CL,
Heymsfield SB, et al. Body composition estimates from
NHANES III bioelectrical impedance data. Int J Obes Relat
Metab Disord. 2002;26:1596–609.
23. Freedman LS, Graubard BI, Schatzkin A. Statistical validation of
24. VanderWeele TJ, Vansteelandt S. Odds ratios for mediation
analysis for a dichotomous outcome. Am J Epidemiol. 2010;172:
25. StataCorp. Stata Statistical Software: release 11. College Station:
StataCorp LP; 2009.
26. Gallo MF, Lopez LM, Grimes DA, Schulz KF, Helmerhorst FM.
Combination contraceptives: effects on weight. Cochrane Data-
base Syst Rev. 2011;(9). doi:10.1002/14651858.CD003987.pub4.
27. Straub RH. The complex role of estrogens in inflammation.
Endocr Rev. 2007;28:521–74.
28. Klein SL, Jedlicka A, Pekosz A. The Xs and Y of immune
responses to viral vaccines. Lancet Infect Dis. 2010;10:338–49.
29. Pfeilschifter J, Ko ¨ditz R, Pfohl M, Schatz H. Changes in proin-
flammatory cytokine activity after menopause. Endocr Rev.
30. Rogers AER. The effect of 17beta-estradiol on production of
cytokines in cultures of peripheral blood. Bone. 2001;29:30–4.
31. Lacort M, Leal AM, Liza M, Martı ´n C, Martı ´nez R, Ruiz-
Larrea MB. Protective effect of estrogens and catecholestro-
gens against peroxidative membrane damage in vitro. Lipids.
32. Omoya T, Shimizu I, Zhou Y, Okamura Y, Inoue H, Lu G, et al.
Effects of idoxifene and estradiol on NF-kappaB activation in
cultured rat hepatocytes undergoing oxidative stress. Liver.
33. Caprio M, Zennaro MC, Feve B, Mammi C, Fabbri A, Rosano G.
Potential role of progestogens in the control of adipose tissue and
salt sensitivity via interaction with the mineralocorticoid receptor.
34. Lacasa D, Le Liepvre X, Ferre P, Dugail I. Progesterone stimu-
lates adipocyte determination and differentiation 1/sterol regula-
tory element-binding protein 1c gene expression. Potential
mechanism for the lipogenic effect of progesterone in adipose
tissue. J Biol Chem. 2001;276:11512–6.
chronic diseases.Stat Med.
1158 J Gastroenterol (2013) 48:1151–1159
35. Kjeld JM, Puah CM, Joplin GF. Changed levels of endogenous
sex steroids in women on oral contraceptives. Br Med J.
36. Gaspard UJ, Romus MA, Gillain D, Duvivier J, Demey-Ponsart
E, Franchimont P. Plasma hormone levels in women receiving
new oral contraceptives containing ethinyl estradiol plus levo-
norgestrel or desogestrel. Contraception. 1983;27:577–90.
37. Caldwell SH, Harris DM, Patrie JT, Hespenheide EE. Is NASH
underdiagnosed among African Americans? Am J Gastroenterol.
38. Solga SF, Clark JM, Alkhuraishi AR, Torbenson M, Tabesh A,
Schweitzer M, et al. Race and comorbid factors predict nonal-
coholic fatty liver disease histopathology in severely obese
patients. Surg Obes Relat Dis. 2005;1:6–11.
39. Conway JM, Yanovski SZ, Avila NA, Hubbard VS. Visceral
adipose tissue differences in black and white women. Am J Clin
40. Lovejoy JC, de la Bretonne JA, Klemperer M, Tulley R.
Abdominal fat distribution and metabolic risk factors: effects of
race. Metabolism. 1996;45:1119–24.
41. Randolph JF Jr, Sowers M, Gold EB, Mohr BA, Luborsky J,
Santoro N, et al. Reproductive hormones in the early menopausal
transition: relationship to ethnicity, body size, and menopausal
status. J Clin Endocrinol Metab. 2003;88:1516–22.
42. Randolph JF Jr, Sowers M, Bondarenko IV, Harlow SD, Lubor-
sky JL, Little RJ. Change in estradiol and follicle-stimulating
hormone across the early menopausal transition: effects of eth-
nicity and age. J Clin Endocrinol Metab. 2004;89:1555–61.
43. Ukkola O, Gagnon J, Rankinen T, Thompson PA, Hong Y, Leon
AS, et al. Age, body mass index, race and other determinants of
steroid hormone variability: the HERITAGE Family Study. Eur J
J Gastroenterol (2013) 48:1151–11591159