Nilsson M, Johnsen R, Ye W, Hveem K, Lagergren J. Obesity and Estrogen as Risk Factors for Gastroesophageal Reflux Symptoms. JAMA. 2003;290(1):66-72. doi:10.1001/jama.290.1.66
Author Affiliations: Department of Surgery, Karolinska Institutet, Karolinska Hospital (Drs Nilsson and Lagergren), and Department of Medical Epidemiology and Biostatistics, Karolinska Institutet (Drs Nilsson, Ye, and Lagergren), Stockholm, Sweden; Department of Community Medicine and General Practice, Norwegian University of Science and Technology, Trondheim, Norway (Dr Johnsen); and Department of Medicine, Levanger Hospital, Levanger, Norway, and HUNT Research Centre, Verdal, Norway (Dr Hveem).
Context Gastroesophageal reflux and obesity are both increasing in prevalence.
The scientific evidence for an association between these conditions is sparse
and contradictory. A difference between sexes concerning this relation has
Objective To evaluate the relation between body mass and gastroesophageal reflux
symptoms and determine how this relation is influenced by female sex hormones.
Design Population-based, cross-sectional, case-control study.
Setting Two consecutive public health surveys within the county of Nord-Trondelag,
Norway, conducted in 1984-1986 and 1995-1997.
Participants Among 65 363 adult participants in the second survey, 3113 individuals
who reported severe heartburn or regurgitation during the last 12 months were
defined as cases, whereas 39 872 persons without reflux symptoms were
defined as controls.
Main Outcome Measure Risk of reflux, estimated using multivariate logistic regression, with
odds ratios (ORs) and 95% confidence intervals (CIs) as measures of association.
Results There was a dose-response association between increasing body mass index
(BMI) and reflux symptoms in both sexes (P for trend
<.001), with a significantly stronger association in women (P<.001). Compared with those with a BMI less than 25, the risk of
reflux was increased significantly among severely obese (BMI >35) men(OR,
3.3; 95% CI, 2.4-4.7) and women (OR, 6.3; 95% CI, 4.9-8.0). The association
between BMI and reflux symptoms was stronger among premenopausal women compared
with postmenopausal women (P<.001), although use
of postmenopausal hormone therapy increased the strength of the association
(P<.001). Reduction in BMI was associated with
decreased risk of reflux symptoms.
Conclusions There is a significant association between body mass and symptoms of
gastroesophageal reflux. The association is stronger among women, especially
premenopausally, and use of hormone therapy strengthens the association, suggesting
that estrogens may play an important role in the etiology of reflux disease.
Gastroesophageal reflux disease and obesity are both highly prevalent
in Western societies, and the occurrence of these conditions is rapidly increasing.1- 7 Reflux
has an adverse impact on quality of life,8 and
the costs for long-term antireflux medication are high.9 Reflux
symptoms10- 12 and
obesity13,14 are strong and independent
risk factors for esophageal adenocarcinoma, a cancer that has increased significantly
in incidence during recent decades.15,16
The relation between body mass and gastroesophageal reflux symptoms
remains uncertain, and valid data to assess this proposed association are
limited. In our previous study, we demonstrated a strong and dose-dependent
association between increasing body mass and endoscopically verified esophagitis
in women, but no association was found in men.17 Moreover,
the positive association among women seemed to be augmented by postmenopausal
hormone therapy, suggesting a role of female sex hormones in the etiology
of reflux disease.17 To examine the relation
between body mass and reflux symptoms and evaluate the influence of female
sex hormones, we conducted a large population-based, cross-sectional, case-control
study of body mass index (BMI) and hormone therapy with respect to the risk
of reflux within a defined cohort.
In the Norwegian County of Nord-Trondelag, 2 extensive public health
surveys have been conducted during recent decades. The first, Helseundersokelsen
i Nord-Trondelag 1 (HUNT 1), was performed in 1984-1986 and included 74 599
individuals, representing 88.1% of the entire adult population (from the year
the patient turned 20 years or older). The second survey, HUNT 2, conducted
in 1995-1997, included 65 363 individuals, representing 71.2% of the
adult population. A total of 47 556 individuals, representing 72.8% of
all persons included in HUNT 2, participated in both surveys.
At their local health centers, all participants completed extensive
written questionnaires that covered a wide variety of exposures. A total of
813 variables were included in both surveys together. Data were collected
on defined disorders and diseases, lifestyle factors, behavioral habits, work
and employment information, use of medications, and aspects of psychosocial
well-being. Furthermore, all participants underwent certain physical examinations,
including assessment of body weight and height, and BMI18 (weight
in kilograms divided by the square of height in meters) was calculated based
on objective measurements. Information on other exposures of interest, including
present or previous postmenopausal hormone therapy, menstrual status, and
previous hysterectomy, was gathered in the questionnaire.
The outcome was defined as severe symptoms of reflux (eg, recurrent
heartburn or regurgitation) and was assessed in HUNT 2 only. Participants
answered a question regarding whether they had experienced heartburn or regurgitation
during the past 12 months and, if so, if the symptoms were minor or severe.
Among the 58 596 persons (90%) in the HUNT 2 survey who answered this
question, 40 210 (69%) reported that they had had no reflux symptoms,
15 233 (26%) had had minor symptoms, and 3153 (5%) had experienced severe
symptoms of reflux. The 3153 individuals who reported severe symptoms were
selected to represent the case group and the 40 210 persons without reflux
symptoms were selected as controls. The 15 233 persons with minor symptoms
were analyzed separately, because the symptom pattern in this group was considered
to be heterogeneous, with an increased risk of misclassification of the outcome,
compared with the group reporting severe symptoms.
Heartburn and acid regurgitation are the cardinal symptoms of gastroesophageal
reflux disease, and the use of questionnaires to assess these symptoms is
well validated as a reliable measurement of the true occurrence of reflux.19- 22 To
further evaluate the outcome among persons who reported different levels of
reflux symptoms in the HUNT 2 survey, we conducted a separate validation study.
The question that covered reflux symptoms used in HUNT 2 was compared with
a more extensive, validated questionnaire that covered the frequency and duration
of reflux symptoms, the occurrence of nightly reflux symptoms, the use and
effect of specified antireflux medications, and the effect of the reflux symptoms
on everyday life. A total of 1102 outpatients at general practices in Nord-Trondelag,
the community hospital of Levanger in Nord-Trondelag, and the Karolinska Hospital
in Stockholm, Sweden, were included in the validation study.
The cutoff points for BMI were predetermined and based on the World
Health Organization classification of overweight and obesity.23 A
BMI value between 25 and 30 is defined as overweight, a BMI greater than 30
as obesity, and a BMI greater than 35 as severe obesity. All persons with
a BMI value less than 25 (normal) constituted the reference group in comparisons
between BMI levels. Odds ratios (ORs) and their 95% confidence intervals (CIs),
derived from unconditional logistic regression, were used to assess the association
between BMI and the risk of reflux.24 Linear
trend of the association was tested in a multivariate model by treating categorical
variables as continuous.
Potential confounding effects of age (in 10-year intervals), tobacco
smoking (years of daily smoking subdivided into <1 year, 1-10 years, and
>10 years), alcohol drinking (number of alcoholic beverages consumed during
the 2 weeks preceding data collection, subdivided into 0, 1-4 drinks, 5-10
drinks, and >10 drinks), asthma medication (months of daily use during the
last 12 months), antihypertensive medication (present use), diabetes mellitus
(known diagnosis), coffee use (cups per day), tea use (cups per day), table
salt use (extra salt on regular meals), and dietary fiber intake (predominantly
consumed type of bread classified in terms of dietary fiber content) were
controlled by introducing these variables individually into the model. For
asthma medication, antihypertensive medication, coffee and tea use, and dietary
fiber in bread, data were available from HUNT 2 only. For the other potential
confounders, data were available from both surveys. Whenever possible, data
from HUNT 1 or HUNT 2 for lifetime exposure were used rather than cross-sectional
HUNT 2 data to reduce the risk of reversed causality (ie, that the level of
exposure is affected by the occurrence of reflux symptoms). Moreover, nonspecific
gastrointestinal symptoms, such as nausea (during last 12 months), constipation
(minor or severe during last 12 months), and diarrhea (minor or severe during
last 12 months) were tested as outcome variables in the logistic regression
model (ie, as control symptoms for reflux).
The potential interaction effect (ie, effect modification) between body
mass and hormone therapy was tested by introducing a cross-product term, representing
the interaction between the 2 variables, into the model. To avoid confounding
effects from pregnancy, women who reported that they were pregnant at the
time of the HUNT 2 reflux outcome assessment were excluded from further analysis.
To investigate differences in the risk of reflux related to BMI between premenopausal
and postmenopausal women, data from female study participants were stratified
according to menstrual status and then entered into the multivariate model.
The effects of weight loss and weight gain were assessed by entering the net
change in BMI between the HUNT 1 and HUNT 2 surveys into the model, with adjustment
for baseline BMI (ie, the BMI measured in the HUNT 1 survey). Informed consent
was obtained from all participants in the study. The study was approved by
the Regional Committee for Medical Research Ethics, Region IV, Norway.
In total, the study evaluated 43 363 subjects (20 369 men
and 22 994 women). After excluding 378 pregnant women, 22 616 women
remained for further analysis. Some general characteristics of the participants
are presented in Table 1. The
mean age of the participants was 52 years among the 3113 cases and 48 years
among the 39 872 controls. The age differences between sexes were minor
and not significant. Mean BMI was 28.1 among cases and 25.8 among controls.
Among female cases, 14.8% had ever used hormone therapy, whereas the corresponding
frequency was 9.6% among the female controls.
Among men, a moderate and dose-dependent association between increasing
BMI and reflux symptoms was observed in the multivariate analysis (P for trend <.001) (Table 2).
Severely obese men (BMI >35) demonstrated a more than 3-fold increase in risk
of reflux symptoms (OR, 3.3; 95% CI, 2.4-4.7) compared with men of normal
weight (BMI <25). The corresponding analysis among severely obese (BMI
>35) women revealed a similarly dose-dependent but stronger association compared
with that observed in men (P<.001) (Table 2), with a more than 6-fold increase in risk of reflux symptoms
(OR, 6.3; 95% CI, 4.9-8.0) compared with women of normal weight (BMI <25).
Adjustment for age only revealed no major differences compared with the multivariate
analyses, indicating lack of important confounding by the variables that were
tested (data not shown).
The association between BMI and reflux symptoms was stronger in severely
obese premenopausal women than in severely obese postmenopausal women (P<.001) (Table 2)
(OR, 6.8; 95% CI, 4.7-9.7; and OR, 4.2; 95% CI, 3.2-5.5; respectively) compared
with women of normal weight with the same menstrual status.
Among women who had ever (presently or previously) been treated with
hormone therapy, there was a strong and dose-dependent increase in the risk
of reflux symptoms, especially in the highest BMI intervals, representing
mainly an effect modification (P<.001) of the
BMI effect on risk of reflux symptoms (Table 3). The separate effect (independent of obesity) of hormone
therapy was tested among women of normal weight (BMI <25), and a tendency
toward increased risk of reflux among women who ever received hormone therapy
was found but was not statistically significant (OR, 1.3; 95% CI, 0.9-1.8).
In the group of women treated with hormone therapy after hysterectomy (ie,
women without an endometrium to protect who hence received estrogens only
[without gestagens]), the risk of reflux in normal-weight women was increased
significantly (OR, 2.3; 95% CI, 1.1-4.8) (Table 3).
The risk of reflux symptoms was increased among women with BMI greater
than 35 currently using hormone therapy and among women in the same BMI category
with previous hormone therapy. The highest risk of reflux symptoms observed
in our study was among women treated with estrogen-only hormone therapy (ie,
previous hysterectomy) with a BMI greater than 35, although the number of
women in these groups was small (Table 3). In general, the age-adjusted estimates did not differ importantly
from the estimates from multivariate models (data not shown).
For the 72.8% of the individuals who participated in both surveys, we
repeated the analyses using BMI data from the first survey (ie, approximately
a decade before the assessment of reflux symptoms). A weak association between
increasing body mass and reflux symptoms was found among moderately obese
men (BMI 30-35), with a 40% increase in risk of reflux symptoms (OR, 1.4;
95% CI, 1.1-1.9) compared with men of normal weight (BMI <25). Among severely
obese men (BMI >35), the association was not statistically significant (OR,
1.8; 95% CI, 0.9-3.5).
In women, the association was stronger, with a significant increase
in the risk of reflux symptoms in all 3 categories of overweight and obesity
compared with those with a BMI less than 25. Overweight nonpregnant women
(BMI 25-30) had an increased risk of reflux symptoms compared with participants
with a BMI less than 25 (OR, 2.1; 95% CI, 1.7-2.4). In obese women (BMI 30-35)
the risk of reflux symptoms was increased compared with women of normal weight
(OR, 3.2; 95% CI, 2.5-4.1), and the corresponding risk was similarly increased
in severely obese women (BMI >35) (OR, 2.5; 95% CI, 1.7-4.0). The analyses
concerning exposure to hormone therapy based on data from HUNT 1 are similar
to the results from HUNT 2, although the strength of the associations are
somewhat diluted. Both analyses show strong and dose-dependent increases in
risk of reflux symptoms among hormone therapy users. Similar to the results
from HUNT 2, the strongest association was among severely obese women (BMI
>35) taking estrogen-only hormone therapy.
Participants in both surveys were evaluated for the association between
weight change, identified during the time interval between the 2 surveys,
and risk of reflux (Table 4).
The risk of reflux was dose-dependently greater with increasing net BMI gain.
In the group that gained more than 3.5 BMI units, the risk of reflux symptoms
was increased (OR, 2.7; 95% CI, 2.3-3.2) compared with persons with stable
BMI, whereas the risk of reflux symptoms was decreased among persons who lost
more than 3.5 BMI units (OR, 0.6; 95% CI, 0.4-0.9).
Sex-stratified analyses of the relation between BMI and reflux symptoms
and analyses of the influence of hormone therapy were also performed using
the 15 233 study participants who reported minor reflux symptoms (omitted
from the main analyses because of suspected heterogeneity of symptoms) as
the case group, instead of persons who reported severe symptoms. The results
of these analyses were similar, except with lower point estimates and weaker
trends, compared with the analyses using severe reflux outcome for case classification
(data not shown). In analyses of the association between BMI and hormone therapy
and control symptoms for reflux, the risks of nausea, constipation, and diarrhea
were not significantly affected by either variations in BMI or use of hormone
therapy (data not shown).
In the validation study, which comprises a total of 1102 participants,
103 (9.6%) reported severe reflux symptoms of heartburn or regurgitation during
the past 12 months. In this group (which corresponded to our case group with
severe reflux symptoms), 72% reported heartburn or regurgitation that occurred
at least daily or were taking antireflux medication daily, 23% had heartburn
or regurgitation one or several times per week, and 5% had symptoms less frequently
than once weekly. Hence, 95% of the group corresponding to our case group
had experienced reflux symptoms at least once per week. Among the 280 participants
(25.4%) who reported minor symptoms of heartburn or acid regurgitation during
the past 12 months (corresponding to the group excluded from the main analyses),
15% had daily symptoms or had used antireflux medication on a daily basis,
10% had symptoms at least once per week, and 75% had symptoms less frequently
than once a week. Among those reporting severe reflux symptoms (corresponding
to the case group in the main analyses), the specificity for symptoms that
occurred at least once per week was 99.5% and the corresponding sensitivity
This study demonstrates a strong and dose-dependent association between
increasing body mass and symptomatic reflux in women and a moderate association
among men. The association was stronger among premenopausal women. There was
a weak association between hormone therapy and reflux in women of normal weight.
With increasing body mass, the association between hormone therapy and reflux
became increasingly stronger, suggesting that hormone therapy is an effect
modifier of the association between body mass and reflux. The study also reveals
that weight loss is associated with reduced risk of reflux symptoms.
Strengths of the study include the population-based design with high
participation rates, reducing the risk of selection bias. The large sample
size decreased the risk of chance findings and facilitated extensive subgroup
analyses. The wide range of exposure data made it possible to adjust for potential
A weakness of the study is that reflux was assessed only in the HUNT
2 survey, which prevented us from assessing new cases of reflux that occurred
between the 2 surveys (ie, truly incident cases). This cross-sectional design
theoretically opens the field for reversed causality. However, the consistently
strong and dose-dependent associations among body mass, hormone therapy, and
reflux symptoms, together with the biological implausibility of reflux causing
obesity and the use of hormone therapy, provide reassurance against reversed
Another possible weakness is that reflux was assessed by reflux symptoms
only, with a potential risk of misclassification of the outcome. This is of
less concern, however, because the symptoms used for outcome assessment, heartburn
and regurgitation, are well validated as representing true reflux disease.19- 22 Furthermore,
we validated our question concerning reflux symptoms against a more extensive,
previously used reflux symptom questionnaire.11 The
usual definition of reflux disease, based on symptom evaluation, is a frequency
of symptoms of once per week or more. In the validation study, 95% of the
group corresponding to our case group had reflux symptoms that fulfilled this
definition, confirming our outcome assessment to have high specificity (99.5%).
Moreover, any misclassification of reflux would dilute the associations and
could not explain the positive associations. The fact that the analyses of
the participants who reported minor reflux symptoms revealed similar but weaker
patterns of associations compared with the main analyses of persons with severe
reflux symptoms further supports the credibility of our results.
All analyses used BMI data from the HUNT 1 and HUNT 2 surveys. Results
were similar, revealing the same patterns of associations, but in data from
HUNT 1 the point estimates were generally weaker. This difference between
the surveys is probably because the relation between BMI and reflux is more
dependent on current than previous BMI. In this respect, current, cross-sectional
BMI data may be more relevant to reflux outcome than previous BMI.
Previous scientific evidence concerning the relation between body mass
and reflux is ambiguous. Three population-based, epidemiological studies9,25,26 have evaluated the
association between body mass and reflux symptoms. In our previous cross-sectional
study of reflux symptoms,25 we found no association.
In light of the data presented herein, that might be explained by the low
percentage (17%) of women included and that BMI was assessed retrospectively
at least 20 years before the collection of data. Previous (as opposed to present)
exposure to high BMI levels diluted the results in the present study, particularly
among men. This could explain the negative result in our previous study.25 In a cross-sectional study, Locke et al9 identified
a moderate association, but no separate analysis of men and women was presented.
In a cohort design, Ruhl and Everhart26 demonstrated
a weak association between BMI and hospitalization for reflux diagnoses. Our
previous population-based study examining the relation between body mass and
endoscopically verified that reflux esophagitis demonstrated a strong and
dose-dependent association between increasing body mass and reflux esophagitis
in women, augmented by hormone therapy, whereas no association was found in
men.17 The difference between our previous
study and the present study, concerning the association among men, might be
explained by lack of precision in the previous study.
The findings of our previous report17 led
us to formulate the hypothesis that estrogens might be involved in the pathogenesis
of reflux disease among obese women. The concentration of sex hormone–binding
globulin is lower among obese women than normal-weight women, resulting in
a larger proportion of unbound, active estradiol. Moreover, obese women also
have increased synthesis of estrone in the fatty tissue.27- 29 Reflux
symptoms during pregnancy usually commence during the first trimester30- 33 because
of a predominantly hormonally mediated pathogenesis, and women taking sequential
oral contraceptives have reduced lower esophageal sphincter tone, facilitating
reflux.34 Moreover, estrogen increases nitric
oxide synthesis, the predominant relaxing transmitter substance of the lower
esophageal sphincter,35- 39 resulting
in smooth muscle relaxation in animal models40 and
humans.41 Therefore, we hypothesize that the
association between estrogens and reflux might be conveyed by a nitric oxide–mediated
reduction of smooth muscle tone in the lower esophageal sphincter. Findings
from a randomized, double-blind, crossover, placebo-controlled study42 showing that nitric oxide synthesis increased with
increasing BMI in postmenopausal women with estrogen treatment support this
hypothesis. Our finding that hormone therapy increases the risk of reflux
with increasing degrees of obesity supports a role of female sex hormones
in the etiology of reflux disease. Furthermore, the finding that stratification
for estrogen-only hormone therapy is the only significant effect of hormone
therapy on reflux risk among normal-weight women (BMI <25) and the strongest
effect among severely obese women (BMI >35) suggests that the hormone therapy
effect is mediated by estrogens and not gestagens.
Stratification between present and previous use of hormone therapy reveals
a stronger effect modification of the BMI effect on reflux symptoms among
previous hormone therapy users than among present users. This finding might
be explained by selection bias in that individuals who had reflux symptoms
before hormone therapy might have had exacerbation of symptoms during present
hormone therapy and therefore may have stopped treatment. Such bias does not
affect the "ever hormone therapy" group as a whole, because it only affects
the distribution between the present and previous groups. The finding that
weight loss, manifested as a reduction in BMI between the 2 surveys, was associated
with a significantly decreased risk of reflux symptoms when compared with
individuals with stable BMI supports the role of weight loss in the clinical
management of patients with reflux symptoms.
In conclusion, our large, population-based study provides evidence of
a dose-dependent association between increasing body mass and the risk of
gastroesophageal reflux symptoms. Hormone therapy was a weak risk factor for
reflux symptoms but increased the strength of the association between body
mass and reflux, an effect that was more pronounced with estrogen-only treatment.
Weight reduction is associated with reduced risk of reflux.