Objective: Excess body mass is associated with symptoms of gastro-oesophageal reflux disease, and cross-sectional studies suggest an association between body mass index (BMI) and Barrett’s oesophagus. The present study sought prospectively to examine the influence of BMI and other anthropomorphic measures on the risk for Barrett’s oesophagus among women.
Methods: This was a prospective study of 15 861 women who participated in the Nurses’ Health Study, without a history of cancer, who underwent upper gastrointestinal endoscopy for any reason between 1986 and 2004. The main outcome measures were 261 cases of pathologically confirmed specialised intestinal metaplasia within the oesophagus (Barrett’s oesophagus). Self-reported data on weight were collected from biennial questionnaires. Self-reported height was collected in 1976, and self-reported waist and hip circumferences were collected in 1986.
Results: Compared with women with a BMI of 20–24.9 kg/m2, women with a BMI of 25–29.9 had a multivariate OR for Barrett’s oesophagus of 0.92 (95% CI 0.66 to 1.27), women with a BMI ⩾30 had a multivariate OR of 1.52 (95% CI 1.02 to 2.28) and women with a BMI <20 had a multivariate OR of 0.92 (95% CI 0.65 to 1.31). Results were similar when controlling for symptoms of gastro-oesophageal reflux, and among the entire Nurses’ Health Study cohort (n = 93 609) regardless of a history of endoscopy. In contrast, waist-to-hip ratio, waist circumference and height did not appear to be associated with Barrett’s oesophagus.
Conclusions: Obese, but not overweight, women appear to be at increased risk for Barrett’s oesophagus.
Statistics from Altmetric.com
Barrett’s oesophagus is a metaplastic condition resulting from exposure of the oesophagus to refluxed gastric contents, particularly acid and possibly bile.1 2 It is postulated that denuded squamous mucosa is repopulated by columnar cells originating from pluripotent cells in the basal epithelium.3 4 The prevalence of Barrett’s oesophagus among patients undergoing upper gastrointestinal endoscopy (oesophagogastroduodenoscopy (OGD)) in the USA is between 3% and 25%,5 6 7 8 and 0.3% among the general population.9 Barrett’s oesophagus appears to have a male predominance, with a male to female ratio of 1.7–2.0:1,8 10 11 12 and this may explain the paucity of data about women with this condition.
The prevalence of Barrett’s oesophagus among patients with heartburn is higher than those without, supporting the role of acid reflux in this disease.8 13 14 In addition, factors associated with Barrett’s oesophagus include increased duration of reflux symptoms, increased exposure of the oesophagus to refluxed gastric contents, hiatus hernia and a defective lower oesophageal sphincter.15 16 17 Barrett’s oesophagus is also a precursor lesion for oesophageal adenocarcinoma, with an incidence of 0.4–0.5% per patient-year.10 18 19 20 21 22 23 24
Increased body mass index (BMI) has been associated with an excess risk of gastro-oesophageal reflux disease (GORD), including symptoms of heartburn, erosive oesophagitis and oesophageal adenocarcinoma.25 26 27 28 29 30 31 32 The rising prevalence of overweight and obesity in the USA33 may therefore explain some of the increased incidence of Barrett’s oesophagus34 35 and oesophageal adenocarcinoma.36 37 38 39
Indeed, previous studies have suggested an association between BMI and Barrett’s oesophagus. These studies, however, have been limited by their cross-sectional or retrospective case–control designs, and have had predominantly male populations.7 40 41 42 43 44 We sought to clarify the relationship between excess body mass and Barrett’s oesophagus in women further using data collected prospectively as part of the Nurses’ Health Study, a large, ongoing cohort study in which detailed information on weight and other health-related factors has been collected over 30 years. To understand further any association between BMI and Barrett’s oesophagus, we also sought to explore the relative importance of body mass itself, body fat distribution (e.g. waist circumference and waist-to-hip ratio) and height (a marker of growth factor exposure).
The Nurses’ Health Study cohort was established in 1976 when 121 700 female registered nurses, 30–55 years of age, completed a questionnaire about risk factors for cancer and cardiovascular disease. With an overall response rate exceeding 90%, participants have received follow-up questionnaires every 2 years to obtain information about personal habits (including detailed dietary information every 4 years), medical diagnoses and medication use.
Assessment of BMI and other exposures
We determined BMI (the weight in kilograms divided by the square of the height in metres) from measurements of height provided by participants in 1976 and from measurements of weight reported in 1976 and updated every 2 years thereafter. Waist and hip measurements were self-reported in 1986, and subsequently validated among a sample of 140 Nurses’ Health Study cohort members.45 Trained technicians visited these 140 participants twice, approximately 6 months apart, to measure weight and waist and hip circumferences. The correlation between self-report and the average of the technicians’ two measurements was 0.97 for weight, 0.87 for waist circumference, 0.81 for hip circumference and 0.66 for waist-to-hip ratio.
Smoking status, menopausal status, use of postmenopausal hormones and history of cancer or diabetes were assessed in 1976 by self-report and updated every 2 years thereafter. Dietary information was first obtained using a semi-quantitative food frequency questionnaire in 1980, updated in 1984, 1986 and subsequently every 4 years thereafter. The Nurses’ Health Study food frequency questionnaire has ranged in size over the years from 61 to >120 items. This permitted calculations of daily caloric intake and alcohol use. Physical activity was assessed in 1980, 1986, 1988, 1992, 1994, 1996 and 2000. Each activity reported was measured in metabolic equivalent task (MET)-hours per week. One MET represents the energy expended during 1 h of rest. In 1982, 1994 and every 2 years thereafter, participants were asked to indicate if they used histamine type 2 receptor antagonists “regularly in the past 2 years”. Regular use of proton pump inhibitors was asked in 2000, 2002 and 2004. In 2002, Nurses’ Health Study participants were asked if they “ever regularly had heartburn/acid reflux one or more times a week”. Caloric intake, alcohol use, physical activity, menopausal status and GORD symptoms in this cohort have been validated previously.26 45 46 47 48 49
Ascertainment of cases
In 2002 and 2004 Nurses’ Health Study participants were asked if they had ever undergone OGD or been diagnosed with Barrett’s oesophagus. We requested written permission to acquire endoscopy and pathology records from women reporting Barrett’s oesophagus. A study physician, blinded to exposure information, reviewed records to extract information on the initial date of diagnosis of Barrett’s oesophagus, the length of columnar-lined oesophagus seen at endoscopy and the presence or absence of specialised intestinal metaplasia (SIM) documented in biopsies taken from the oesophagus. Our primary case definition included only women with oesophageal SIM of any length. Secondary case definitions included (1) women with SIM and at least 1 cm of columnar-lined oesophagus and (2) any woman with an oesophageal biopsy demonstrating SIM, columnar epithelium or a pathology report simply stating “Barrett’s oesophagus” without a microscopic description.
To verify that failure to report Barrett’s oesophagus was a reliable indication that a participant did not have the condition, we requested written permission to acquire records from 200 randomly selected women who reported an OGD but not Barrett’s oesophagus. After one mailing attempt, we obtained records from 95 women. In none of these instances did the endoscopist suspect Barrett’s oesophagus.
To avoid detection bias, our primary study population was restricted to those women who reported undergoing OGD during the study period (n = 19 005), as only those who underwent endoscopy could be accurately classified as either having or not having Barrett’s oesophagus. We excluded women with cancer (except non-melanoma skin cancer) prior to their index endoscopy (n = 1293), women who never provided height or weight (n = 140), and those with a BMI <15 kg/m2 (n = 1333). We excluded women whose endoscopy occurred prior to 1987 as waist and hip measurements were ascertained in 1986 (n = 3). To avoid misclassification bias, we excluded women who reported Barrett’s oesophagus but for whom review of records failed to support a diagnosis of at least columnar-lined oesophagus (n = 375).
Women were categorised according to categories of BMI (<20, 20–24.9, 25–29.9 and >30 kg/m2) calculated by taking an updated cumulative average of BMI between 1976 and all available questionnaires up to the 2-year cycle prior to the index endoscopy. We used age- and multivariate-adjusted unconditional logistic regression to obtain odds ratios (ORs) and 95% CIs for the risk of Barrett’s oesophagus. Women with a BMI of 20–24.9 kg/m2 served as the reference population. Multivariate models controlled for year of endoscopy in addition to the covariates described above. We performed several secondary analyses, all planned a priori, to verify the robustness of our findings. These included an analysis controlling for frequent GORD symptoms in 2002, an analysis restricted to women reporting frequent GORD symptoms and an analysis using the entire Nurses’ Health Study cohort regardless of history of endoscopy. In this last case, as we had no date of index endoscopy for most cohort members, we modelled data from our dietary baseline (1980) or the earliest relevant questionnaire.
We also conducted a stratified analysis to determine whether the risk associated with excess weight was modified by the use of acid-blocking medications. We performed a test for interaction by entering into the multivariate model the cross-product term of BMI and acid-blocker use, each as a categorical variable. To examine the effect of weight change on the risk of Barrett’s oesophagus, we categorised women according to their difference in weight between 1976 and their index endoscopy. We calculated the ORs for Barrett’s oesophagus among women categorised by quartiles of height, waist-to-hip ratio and waist circumference. When calculating the mean length of Barrett’s oesophagus, reports describing only “tongue(s)” of Barrett’s oesophagus, a “short-segment” Barrett’s oesophagus or an “irregular z-line” were considered to be 0.5 cm in length. Finally, we assessed for an association between BMI and length of Barrett’s oesophagus using linear regression.
Analyses were performed with SAS, version 9.1 (SAS Institute, Cary, North Carolina, USA). All p values are two-sided. The current study was approved by the institutional review boards of Brigham and Women’s Hospital and Boston University Medical Center.
Among 15 861 eligible women, we documented 261 pathologically confirmed cases of SIM, and an additional 80 cases with either columnar epithelium within the oesophagus or simply a pathologist’s diagnosis of “Barrett’s oesophagus” without documentation of SIM. Among the cases with SIM, 20 (8%) were diagnosed prior to 1994, 92 (35%) were diagnosed between 1994 and 1999, and 149 (57%) were diagnosed between 2000 and 2004. As a percentage of endoscopies during each timeframe, SIM was present in 0.6% of endoscopies prior to 1994, 1.6% of endoscopies between 1994 and 1999, and 2.2% of endoscopies between 2000 and 2004. Length was described to some extent by the endoscopist in 330/341 (97%) instances of columnar-lined oesophagus. Among those with segments <1 cm, the endoscopist had described only an “irregular z-line”, a “short-segment Barrett’s oesophagus” or “tongue(s)” of Barrett’s oesophagus. The mean (SD) length of columnar-lined esophagus was 1.9 (2.6) cm, with 74 (22%) having a segment length >3 cm. There appeared to be no association between BMI and length of Barrett’s oesophagus (parameter estimate 0.03; p = 0.31).
Among eligible women, 24% were overweight (BMI 25–29.9 kg/m2) and 9% were obese (BMI >30 kg/m2). Compared with lean women (BMI <25 kg/m2), overweight and obese women (BMI >25 kg/m2) were less likely to be current smokers (p<0.001), to use postmenopausal hormones (p<0.001) or to participate in regular vigorous exercise (p<0.001). Overweight and obese women consumed more calories (p = 0.006), although less alcohol (p<0.001), and were more likely to have diabetes mellitus (p<0.001), experience symptoms of GORD (p<0.001) and use acid-blocking medications (p<0.001; table 1).
BMI and risk for Barrett’s oesophagus
We observed a significantly increased risk for Barrett’s oesophagus among obese, but not overweight, women (table 2). Compared with women having a BMI of 20–24.9 kg/m2, those with a BMI of 25–29.9 kg/m2 had a multivariate OR for Barrett’s oesophagus of 0.92 (95% CI 0.66 to 1.27), while those with a BMI >30 kg/m2 had a multivariate OR of 1.52 (95% CI 1.02 to 2.28). Our findings were similar between age-adjusted univariate and multivariate models, suggesting minimal confounding by the other covariates tested (table 2). The multivariate models presented yielded the same results as more complicated models controlling for a cumulative average of aspirin and non-steroidal anti-inflammatory drug use (data not shown).
We observed similar risks for Barrett’s oesophagus when using a less stringent definition of Barrett’s oesophagus, requiring only the presence of columnar histology in an oesophageal biopsy (table 2). When restricting the definition of Barrett’s oesophagus to SIM of at least 1 cm length, the association between obesity and Barrett’s oesophagus was no longer statistically significant, although the point estimate was similar to our primary endpoint and power was reduced due to a smaller sample size (OR 1.75; 95% CI 0.97 to 3.14).
To minimise detection or selection biases, our primary analysis restricted eligibility to participants who had undergone OGD. Nonetheless, we conducted a secondary analysis that included all participants in the Nurses’ Health Study, regardless of having an OGD (n = 93 609). In this analysis, compared with women having a BMI of 20–24.9, the multivariate OR for Barrett’s oesophagus among obese women (BMI >30 kg/m2) was 1.55 (95% CI 1.04 to 2.29; table 2). In all these analyses, excess risk for Barrett’s oesophagus was only experienced by women with a BMI >30 kg/m2; women with a BMI of 25–29.9 kg/m2 appeared to have the same risk as the reference population.
Controlling for frequent GORD symptoms did not materially alter the observed risks for Barrett’s oesophagus, although the association between obesity and Barrett’s oesophagus was no longer statistically significant. Compared with women having a BMI of 20–24.9 kg/m2, those with a BMI of 25–29.9 kg/m2 had a multivariate OR of 0.89 (95% CI 0.64 to 1.24), while those with a BMI >30 kg/m2 had a multivariate OR for Barrett’s oesophagus of 1.46 (95% CI 0.98 to 2.20). These findings were similar in an analysis restricted only to women who reported frequent GORD symptoms in 2002 and had undergone OGD (n = 9436 total; 224 cases; table 2).
To define the risk for Barrett’s oesophagus further among women with GORD symptoms, we stratified women according to whether or not they reported regular use of acid-blocking medications prior to their endoscopy. Among participants who reported regular use of acid-blocking medications prior to their index endoscopy, obese women had a multivariate OR for Barrett’s oesophagus of 1.94 (95% CI 0.94 to 4.01) compared with those having a BMI of 20–24.9 kg/m2. Although no longer statistically significant, the point estimate was similar to our primary endpoint and power was reduced due to a smaller sample size. Formal testing revealed no evidence for an interaction between BMI and acid-blocker use (p = 0.82).
We found no association between weight change and Barrett’s oesophagus (table 3). Compared with women whose weight remained stable between 1976 and their index endoscopy, women who gained >22.6 kg had a multivariate OR for Barrett’s oesophagus of 1.08 (95% CI 0.64 to 1.83). This was similar when restricting the analysis to women with a BMI <25 kg/m2 in 1976 (table 3).
Waist circumference, waist-to-hip ratio and height
To evaluate whether other anthropomorphic measurements are also associated with Barrett’s oesophagus, we categorised participants according to quartiles of waist-to-hip ratio, waist circumference and height. We observed no significant association between these measures and Barrett’s oesophagus (table 4). Simultaneous inclusion of both BMI and waist-to-hip ratio in the same model failed to attenuate the ORs appreciably for either variable (data not shown). A secondary analysis using clinically meaningful measures of waist circumference and waist-to-hip ratios50 also failed to demonstrate any significant association between these measures and Barrett’s oesophagus. For example, compared with those women with a waist circumference of <80 cm, the multivariate ORs for those with a waist circumference of 80 and 88 cm were 0.93 (95% CI 0.62 to 1.39) and 0.99 (95% CI 0.63 to 1.55), respectively.
We found a positive association between obesity and Barrett’s oesophagus among a large cohort of women who underwent OGD for any reason. The observed risk was similar when restricting our analyses to women with frequent GORD symptoms and those who regularly took acid-blocking medications prior to their index endoscopy. Controlling for other known or suspected risk factors for Barrett’s oesophagus did not alter these findings. This risk for Barrett’s oesophagus appears limited to obesity, as overweight women had a risk similar to women with a BMI of 20–24.9 kg/m2. Waist-to-hip ratio, waist circumference and height did not appear to be strongly associated with Barrett’s oesophagus in this cohort. We also noted an increasing temporal trend in the percentage of endoscopies during which Barrett’s oesophagus was diagnosed. However, we were unable to determine whether this was due to an increased prevalence of disease or an increased awareness by endoscopists.
Previous studies have documented an association between BMI and GORD. The risk for GORD symptoms rises progressively with increasing BMI, even among normal weight individuals.26 Erosive oesophagitis and oesophageal adenocarcinoma have also been prospectively associated with BMI.30 32 51 Various mechanisms have been proposed to explain these associations, including increased numbers of transient lower oesophageal sphincter relaxations,52 increased gastro-oesophageal pressure gradients,53 an increased prevalence of hiatus hernia54 and excess production of endogenous oestrogen.27 48
It is therefore reasonable to consider that the observed risk between BMI and Barrett’s oesophagus is primarily mediated through increased gastro-oesophageal reflux. However, the epidemiology of GORD suggests a more complex story. While men and women report GORD symptoms with equal frequency,55 men are approximately twice as likely as women to have Barrett’s oesophagus,56 and 5–8 times as likely to develop oesophageal adenocarcinoma.57
In the present study, our results were similar even after accounting for GORD symptoms and the previous use of acid-blocking medications, suggesting obesity may play a role in Barrett’s metaplasia beyond the promotion of GORD. This conclusion differs from that of a recent meta-analysis of previously unpublished data culled from studies designed to answer various questions about either oesophageal adenocarcinoma or Barrett’s oesophagus.58 The authors found no association between BMI and Barrett’s oesophagus when Barrett’s cases were compared with controls with oesophagitis or positive oesophageal pH monitoring. In our study, we attempted to address this by performing a secondary analysis limiting the entire cohort to those with frequent GORD symptoms. In this case, the association between obesity (defined by a cumulative average BMI, assessed prospectively over many years) and Barrett’s oesophagus remained, and was consistent with the risk observed among the entire endoscopy cohort and the entire Nurses’ Health Study population.
The difference between our study and the meta-analysis may reflect the choice of control population. For example, if oesophagitis itself is associated with obesity, the use of patients with oesophagitis as controls may yield a null result, even though obesity could be associated with both oesophagitis and Barrett’s oesophagus.59 However, we also acknowledge that GORD symptoms were assessed among the entire Nurses’ Health Study population in 2002 only. We therefore relied on participants’ recollection of prior symptoms throughout their life, making our GORD symptom analyses cross-sectional in design. Nonetheless, we prospectively examined the association between obesity and Barrett’s oesophagus among regular users of acid-suppressing medications, finding a multivariate OR for Barrett’s oesophagus of 1.94 (95% CI 0.94 to 4.01). While not statistically significant, the risk estimate suggests that obesity itself is independently associated with Barrett’s oesophagus.
It is intriguing to consider gender differences in body fat distribution when studying the association between BMI and Barrett’s oesophagus. For example, overweight and obese men tend to have more centralised, visceral fat, while women deposit more body fat in their subcutaneous tissues. This may partly explain why measures of fat distribution appeared more strongly associated with Barrett’s oesophagus than BMI among predominantly male populations,40 42 60 while BMI itself may be more important among women. Visceral fat, as observed to a greater extent among men than women, is more active in the production of leptin and adiponectin, two adipokines that have been implicated in the pathogenesis of Barrett’s oesophagus.61 62 63 64 This may also help clarify why both overweight and obese men experience excess risk for Barrett’s oesophagus,40 42 while we observed excess risk only among obese women.
The strengths of our study include its prospective collection of exposure variables (thereby avoiding the potential biases inherent in cross-sectional and retrospective analyses), repeated assessments of BMI, detailed data on potential confounders and a large number of cases. Nevertheless, we acknowledge certain limitations. We relied on self-reported measures of height, weight, and waist and hip circumference. However, these measures have been validated previously in our cohort.45 Furthermore, our use of an updated cumulative average of BMI as our primary exposure limits the effects of year to year variation and inconsistencies in reporting accuracy. Waist and hip measurements were assessed only in 1986, and therefore were not updated. However, as all participants had their index endoscopy after 1986, we were able to maintain a prospective design.
Our definition of Barrett’s oesophagus was based upon review of endoscopy and pathology reports, but did not include central review of pathology specimens, potentially resulting in misclassification bias. However, in a recent study of BMI and Barrett’s oesophagus.40 independent pathologist review confirmed SIM in 91% of 616 cases identified in a manner similar to ours.65 We also employed three different definitions of Barrett’s oesophagus to verify that our findings were robust. The first, requiring the presence of SIM, is the standard definition of Barrett’s oesophagus used in the USA.66 The second definition required the presence of at least 1 cm of columnar epithelium with documented SIM. This should have increased the specificity of our definition by excluding cases of intestinal metaplasia of the gastro-oesophageal junction, a potentially distinct entity.67 The third definition of Barrett’s oesophagus required only the presence of columnar histology, as currently used by the British Society of Gastroenterology.68
In some of our secondary analyses, the association between obesity and Barrett’s oesophagus was no longer statistically significant. However, the ORs in all cases were similar to our primary analysis, suggesting power was reduced due to a smaller sample size.
In summary, we found that obese, but not overweight, women are at increased risk for Barrett’s oesophagus. This risk did not appear to be explained solely by body fat distribution. These findings differ from observations among predominantly male cohorts wherein central adiposity may account for much of the risk. As the predominant pattern of body fat distribution differs between men and women (ie, central vs peripheral, respectively) this may account for the skewed male to female preponderance of Barrett’s oesophagus. Further studies are needed to determine if significant weight loss can decrease the risk for Barrett’s oesophagus.
Funding This work was supported by the National Cancer Institute (CA087969; CA107412 to ATC) and the National Institutes of Diabetes and Digestive and Kidney Diseases (DK070706 to BCJ). ATC is a recipient of the Damon Runyon Cancer Research Foundation Clinical Investigator Award. The sponsors had no role in the design and conduct of the study; collection, management, analyses and interpretation of the data; or preparation, review or approval of the manuscript.
Competing interests None.
Provenance and Peer review Not commissioned; externally peer reviewed.
See Commentary, p 1437
Ethics approval The study was approved by the institutional review boards of Brigham and Women’s Hospital and Boston University Medical Center.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.