RSS Feeds
Meta-Analysis: Obesity and the Risk for Gastroesophageal Reflux Disease and Its Complications
- Howard Hampel, MD, PhD;
- Neena S. Abraham, MD, MSc(Epi); and
- Hashem B. El-Serag, MD, MPH
- From the Michael E. DeBakey Veterans Affairs Medical Center, Baylor College of Medicine, Houston, Texas.
Abstract
Background: The association of body mass index and gastroesophageal reflux disease (GERD), including its complications (esophagitis, Barrett esophagus, and esophageal adenocarcinoma), is unclear.
Purpose: To conduct a systematic review and meta-analysis to estimate the magnitude and determinants of an association between obesity and GERD symptoms, erosive esophagitis, Barrett esophagus, and adenocarcinoma of the esophagus and of the gastric cardia.
Data Sources: MEDLINE search between 1966 and October 2004 for published full studies.
Study Selection: Studies that provided risk estimates and met criteria on defining exposure and reporting outcomes and sample size.
Data Extraction: Two investigators independently performed standardized search and data abstraction. Unadjusted and adjusted odds ratios for individual outcomes were obtained or calculated for each study and were pooled by using a random-effects model.
Data Synthesis: Nine studies examined the association of body mass index (BMI) with GERD symptoms. Six of these studies found statistically significant associations. Six of 7 studies found significant associations of BMI with erosive esophagitis, 6 of 7 found significant associations with esophageal adenocarcinoma, and 4 of 6 found significant associations with gastric cardia adenocarcinoma. In data from 8 studies, there was a trend toward a dose–response relationship with an increase in the pooled adjusted odds ratios for GERD symptoms of 1.43 (95% CI, 1.158 to 1.774) for BMI of 25 kg/m2 to 30 kg/m2 and 1.94 (CI, 1.468 to 2.566) for BMI greater than 30 kg/m2. Similarly, the pooled adjusted odds ratios for esophageal adenocarcinoma for BMI of 25 kg/m2 to 30 kg/m2 and BMI greater than 30 kg/m2 were 1.52 (CI, 1.147 to 2.009) and 2.78 (CI, 1.850 to 4.164), respectively.
Limitations: Heterogeneity in the findings was present, although it was mostly in the magnitude of statistically significant positive associations. No studies in this review examined the association between Barrett esophagus and obesity.
Conclusion: Obesity is associated with a statistically significant increase in the risk for GERD symptoms, erosive esophagitis, and esophageal adenocarcinoma. The risk for these disorders seems to progressively increase with increasing weight.
Frequent symptoms of gastroesophageal reflux disease (GERD) affect between 10% and 20% of adults in the United States. The prevalence of GERD-related complications, including erosive esophagitis, Barrett esophagus, and esophageal adenocarcinoma, has been steadily increasing in the United States and western Europe. For instance, hospitalizations with GERD among veterans increased 10-fold from the 1970s to the 1990s (1). Similarly, the incidence of esophageal adenocarcinoma increased 4-fold over the past 20 years (2-5). The reasons for the increase in GERD and its complications are not known. Changes in diet, prescription medication use, smoking, and alcohol intake and the declining prevalence of Helicobacter pylori infection have been proposed (6-11). Studies have also hypothesized that the increasing trend of obesity in western populations has paralleled the increase in esophageal adenocarcinoma and may be an important factor in this change (12-14).
The notion of obesity as a cause of GERD is biologically plausible. Obesity has been associated with increased intra-abdominal pressures (15), impaired gastric emptying (16), decreased lower esophageal sphincter pressure, and increased frequency of transient sphincter relaxation (17, 18), thus leading to increased esophageal acid exposure. However, the epidemiologic evidence linking obesity to GERD and its complications has not been critically reviewed. We aimed to evaluate, quantify, and summarize the association of obesity to GERD and its complications.
Methods
Literature Search
Two investigators independently searched the published English-language literature (through October 2004) by using MEDLINE. Search terms included obesity or body mass or anthroprometry searched with reflux or heartburn, (o)esophagitis, Barrett's or Barretts, and (o)esophageal cancer or (o)esophageal adenocarcinoma. We performed a recursive hand search of cited bibliographies to increase completeness.
Study Selection Criteria
The following inclusion criteria had to be fulfilled: 1) cross-sectional, case–control, or cohort study that permitted assessment of a causal association between overweight or obesity and esophageal disease; 2) clear definition of obesity as defined by a body mass index (BMI) in kg/m2 or height-to-weight ratio; and 3) well-defined outcome of interest that included GERD symptoms defined by using validated symptom score, esophageal erosions defined by endoscopy, and Barrett esophagus or esophageal adenocarcinoma validated by pathology review. We excluded case reports and case series, studies with fewer than 50 case-patients, and studies that did not report risk estimates or raw data to allow independent calculation of these estimates. If a study met the selection criteria except for failure to report risk estimates, we contacted the study authors in an attempt to obtain these data.
Statistical Analysis
We abstracted or calculated odds ratios for cross-sectional and case–control studies and risk ratios for cohort studies as the risk estimates for associations between obesity and each outcome of interest. We pooled the results in 2 different ways. All studies either presented or had sufficient information to allow the calculation of unadjusted odds ratios estimates (and 95% CIs). We subsequently pooled these unadjusted estimates. Because of the inconsistency in reporting adjusted risk estimates, we estimated the degree of confounding produced by uncontrolled variables in studies that did not report adjusted risk estimates by using a factorization of the relative risk by external adjustment (19). Subsequently, we pooled the adjusted estimates. In addition, we displayed the adjusted odds ratios in a tabular form, indicating the variables that were adjusted for. We conducted pooling by using a random-effects model. We examined the weighted pooled risk estimates for 2 recognized categories: overweight and obesity (20, 21). We assessed heterogeneity with a chi-square test and by visual inspection of Forest plots. We performed statistical analyses with Comprehensive Meta-Analysis (Biostat, Englewood, New Jersey) and Stata 8.0 (Stata Corp., College Station, Texas). We examined funnel plots to evaluate for publication bias.
Results
GERD Symptoms
We identified 370 potentially relevant titles. Of these, 9 studies satisfied our inclusion and exclusion criteria (22-30). All studies were cross-sectional examinations of GERD symptoms in randomly selected samples of the general population (Table 1). Only 3 studies (Nilsson and colleagues [(29], Murray and colleagues [28], and Lagergren and colleagues [26]) were conducted for the primary purpose of evaluating the relationship between obesity and GERD.
Of the 9 studies, 6 studies showed a statistically significant association between obesity and GERD (23-25, 28-30), and 3 studies showed no association (22, 26, 27). Eight of the studies presented data to allow for calculation of unadjusted and adjusted odds ratios and 95% CIs for GERD, overweight (BMI, 25 kg/m2 to 30 kg/m2), and obesity (BMI > 30 kg/m2). In these 8 studies, the pooled weighted unadjusted odds ratios for GERD symptoms among overweight and obese persons were 1.5- and 2.0-fold, respectively, greater than normal-weight persons (Figure 1; Appendix Figure 1).
Table 1 presents the adjusted odds ratios. Seven studies adjusted for age and sex. Two studies adjusted for race (24, 27), 3 studies adjusted for nonsteroidal anti-inflammatory drug intake (23, 28, 30), 5 studies adjusted for cigarette smoking (23, 26, 28-30), 1 study adjusted for H. pylori status and socioeconomic status (28), and 1 study adjusted for physical activity (26). None of these adjustments changed the statistical significance or direction of the observed associations. Similarly, adjustments for alcohol (23, 26, 28-30), coffee consumption (23, 28-30), or other dietary factors (such as dietary fiber [29] or total energy intake [26]) resulted in no appreciable changes in the odds ratios. Nilsson and colleagues (29) found that adjustment for postmenopausal hormone therapy in women strengthened the association.
Two studies (Andersen and Jensen [22] and Stanghellini [25]) did not adjust for confounding factors. On the basis of findings from the studies that provided both adjusted and unadjusted odds ratios, we estimated the effect of confounding variables to reduce the unadjusted ratio by a factor of 0.13, which we applied to the 2 studies. The pooled adjusted odds ratios were 1.43 (95% CI, 1.158 to 1.774; P = 0.001) for BMI of 25 kg/m2 to 30 kg/m2 and 1.94 (CI, 1.468 to 2.566; P < 0.001) for BMI greater than 30 kg/m2 (Figure 1). The heterogeneity among the results of these studies was statistically significant (chi-square, 82.1; P < 0.001).
Erosive Esophagitis
Our search identified 111 potentially relevant titles. Of these, 7 studies (4 case–control studies [31-34], 2 cross-sectional studies [35, 36], and 1 cohort study [37]) fulfilled the inclusion and exclusion criteria (Table 2). Three studies (Wilson and colleagues [33], Stene-Larson and colleagues [32], and Nilsson and colleagues [31]) were conducted for the primary purpose of evaluating the relationship between obesity and esophagitis.
We calculated a pooled odds ratio for the 4 case–control and 2 cross-sectional studies, excluding the 1 cohort study by Ruhl and Everhart (37). The pooled unadjusted odds ratio of esophagitis related to BMI of 25 kg/m2 or higher was 1.7-fold greater than that of esophagitis related to BMI less than 25 kg/m2 (Appendix Figure 2). When we excluded the 2 studies from Asia (35, 36) from our analysis, the association between obesity and esophagitis did not appreciably change (pooled unadjusted odds ratio, 1.8 [CI, 1.5 to 2.2]) and there was no statistical heterogeneity (chi-square, 0.15; P > 0.2).
Table 2 shows adjusted odds ratios. Four studies adjusted for potential confounding factors (31, 33, 34, 37). Wilson and colleagues (33) found that the association lost statistical significance after adjustment for hiatal hernia. We estimated from these 4 studies that adjustment for confounding variables decreased the odds ratios by 0.16. We applied this factor to the studies by Stene-Larsen and colleagues (32) and Furukawa and colleagues (36). The pooled adjusted odds ratio for 6 studies was 1.76 (CI, 1.156 to 2.677; P = 0.004) (Figure 2). A heterogeneity test among the studies was also significant (chi-square, 18.9; P = 0.002).
Barrett Esophagus
No studies reporting the association of Barrett esophagus and obesity met the selection criteria of our review.
Esophageal Adenocarcinoma and Adenocarcinoma of the Gastric Cardia
Our search strategy identified 93 titles. Of these, 9 case–control studies met the inclusion and exclusion criteria. Eight studies examined population-based samples, and 1 study examined hospital-based samples (38). Four studies (Chow and colleagues [(39], Ji and colleagues [40], Zhang and colleagues [41], and Lagergren and colleagues [42]) were conducted for the primary purpose of evaluating the relationship between obesity and esophageal adenocarcinoma.
The studies of Chow and colleagues (39), Lagergren and colleagues (42), Vaughan and colleagues (43), and Wu and colleagues (44) examined esophageal adenocarcinoma as a separate category from gastric cardia adenocarcinoma. Cheng and colleagues (45) excluded cancer of the cardia but may have included cancer of the gastroesophageal junction. Brown and colleagues (46) and Zhang and colleagues (41) combined data for adenocarcinomas of the esophagus with the gastric cardia. The 2 studies from China included only cases of gastric cardia adenocarcinoma (38, 40).
Because substantial weight loss at the time of cancer diagnosis was a concern, all studies, except 1 study (38), obtained historical weight information. Among the 3 studies that collected several historical heights and weights (39, 42, 44), the magnitude or direction of the association between BMI at different time points and risk for adenocarcinoma did not statistically significantly differ.
Esophageal Adenocarcinoma
Seven studies examined the association between obesity and esophageal adenocarcinoma (Table 3). Weighted pooling of unadjusted odds ratios (Appendix Figure 3) indicates that the risk for esophageal adenocarcinoma is 2.1 times higher in persons with BMI of 25 kg/m2 or greater than in normal-weight persons. We did not include Zhang and colleagues' study (41) in the pooled odds ratio because the study combined cases of both esophageal and gastric cardia cancer.
All studies adjusted for age and sex, and these were not found to have any appreciable effect on the association. Further adjustments for race (41, 43, 44), smoking (39-46), alcohol consumption (40-43, 45, 46), caloric intake (40-42, 45, 46), history of reflux symptoms (42, 45), or education level (40-44) did not alter the statistical significance or direction of unadjusted associations. The pooled adjusted odds ratio from 6 studies for BMI of 25 kg/m2 or greater was 2.02 (CI, 1.534 to 2.669; P < 0.001). There was a trend toward a dose–response relationship with an increase in pooled adjusted odds ratio for BMI of 25 kg/m2 to 30 kg/m2 and BMI greater than 30 kg/m2 of 1.52 (CI, 1.147 to 2.009; P = 0.004) Figure 3, top) and 2.78 (CI, 1.850 to 4.164; P < 0.001) Figure 3, bottom).
Adenocarcinoma of the Gastric Cardia
Six studies examined the association of obesity and adenocarcinoma of the gastric cardia (Table 3). Four studies showed a statistically significant association (Table 3). Vaughan and colleagues (43) showed an association that did not reach statistical significance. One study from China by Zhang and colleagues (38) showed a statistically significant inverse association. We pooled unadjusted estimates from 5 studies (Figure 3 and Appendix Figure 3) with a modest, weighted pooled odds ratio of 1.5. In general, adjustment resulted in an increase in the odds ratios by a factor of 0.3. We applied this factor to derive adjusted odds ratios in studies that provided only unadjusted estimates. The pooled adjusted odds ratio from all 6 studies was 1.68 (CI, 1.197 to 2.351; P = 0.003). These results had statistically significant heterogeneity (chi-square, 45.3; P < 0.001). However, the 1 study with an inverse association examined BMI 1 year before the cancer diagnosis, whereas other studies used more distant BMI. Excluding that study reduced the heterogeneity (chi-square, 7.6; P = 0.1). Finally, funnel plots were not suggestive of publication bias (not shown).
Discussion
We performed a systematic review of the literature examining the association between BMI and acid-related disorders of the esophagus. The risk for GERD symptoms, erosive esophagitis, or esophageal adenocarcinoma increased with overweight or obesity compared with normal BMI. The association of overweight and obesity with adenocarcinoma of the gastric cardia, a condition less frequently associated with GERD, was weaker and less consistent.
We did not provide a quality score for the individual studies included in our review. However, our inclusion and exclusion criteria ensured a minimum standard of quality. Thus, all studies used acceptable definitions of exposure (BMI) and outcomes of interest (symptoms, endoscopic findings, and pathologic confirmation). Most studies that examined symptoms and cancer were population-based studies with acceptable enrollment rates. However, only 10 of 25 studies stated that the primary purpose of the study was to examine the relationship between obesity and GERD. Another limitation that particularly applies to case–control studies evaluating adenocarcinoma is recall bias, where case-patients may overestimate their historical weight compared with noncancer controls. However, the consistency across studies provides some reassurance against recall bias. To minimize the possibility of recall bias, 3 studies examined patients with squamous-cell carcinoma of the esophagus (39, 42, 43) and 4 studies examined patients with distal gastric adenocarcinoma (38-40, 44) as a second cancer comparison group. These studies found that BMI had either no association or a statistically significant negative association with the cancer controls. Furthermore, in all erosive esophagitis studies, BMI was measured directly.
The heterogeneity test indicated statistically significant degrees of heterogeneity among studies. However, the association between BMI and GERD complications was markedly consistent, and differences in the risk estimates were largely in the magnitude rather than the direction of the association. No studies had a negative association, and the 4 studies with no significant association indicated a trend toward statistical significance. Therefore, the use of pooled odds ratios was appropriate. In the case of esophagitis, the heterogeneity was largely attributed to geographic differences among the study samples. Removing studies that were conducted in Asia eliminated heterogeneity among studies of esophagitis.
Overweight and obesity satisfy several criteria for a causal association with GERD and its complications, including esophagitis and esophageal adenocarcinoma. Our review indicates a consistent modest statistical association, a correct temporal relationship, and possibly a weak dose–response relationship. However, estimates of BMI in most studies were obtained for several years predating the outcomes. The cross-sectional and case–control design of most studies makes it difficult to be emphatic about temporal association, especially for chronic slowly progressive conditions, such as GERD symptoms and erosive esophagitis.
Obese patients may experience extrinsic gastric compression by surrounding adipose tissue, leading to an increase in intragastric pressures and subsequent relaxation of the lower esophageal sphincter (47-49). The findings of manometric studies, however, have been inconsistent, indicating both decreased (50) and normal (17, 51-53) lower esophageal sphincter pressures in obese persons.
Studies have also suggested that rather than obesity, the amount and type of dietary intake, notably fat, are responsible for GERD. Seven studies in our review examined total caloric intake and dietary intake of fiber, fruits and vegetables, or other macronutrients or micronutrients. In all, the effect of BMI on GERD-related disorders was independent of dietary intake, which is further supported by a recent smaller study (54). Moreover, all studies found no consistent association between dietary fat and GERD or esophageal adenocarcinoma.
Obese patients may have an increased risk for hiatal hernia, which has a role in initiating and promoting GERD (55). In our review, 3 studies specifically examined hiatal hernia in relation to obesity and GERD or esophagitis (27, 32, 33). Wilson and colleagues (33) found obesity to be statistically significantly associated with esophagitis. However, after the authors controlled for the presence of hiatal hernia, this association lost statistical significance, implying that hiatal hernia was the mechanism by which obesity leads to esophagitis. Wu and colleagues (44), on the other hand, found no statistically significant association between BMI and hiatal hernia. In a study that examined the association of BMI and grade of esophagitis in persons with esophagitis, El-Serag and Johanson (56) also found that obesity remained an independent risk factor for severe versus mild esophagitis while controlling for the presence of hiatal hernia. The cross-sectional nature of these studies makes it impossible to distinguish temporal associations between obesity and hiatal hernia. In sum, data about hiatal hernia as a cause of increased esophageal reflux in obesity are conflicting.
Humoral factors have also been considered as a mechanism relating obesity to reflux and esophageal adenocarcinoma. Two studies identified in our review conducted by the same group of investigators from Sweden observed that the association of obesity and GERD might be mediated by estrogen (29, 31). The first study reported a statistically significant association between obesity and esophagitis in women, which was potentiated by the use of estrogen in postmenopausal women (31). The second study examined a large population-based cohort and found that obese women had an increased risk for GERD symptoms compared with obese men; the risk was highest in premenopausal women and postmenopausal women receiving estrogen therapy (29). Furukawa and colleagues (36) reported that overweight women older than 70 years of age were the only group to have a statistically significantly increased risk for esophagitis. The authors did not report on estrogen use in these patients. However, several other studies have found that the obesity-related increased risk for esophageal disease was neither confounded nor modified by sex (27, 56).
Some evidence suggests that a change in weight can also affect the risk for GERD and its complications. For example, Nilsson and colleagues (29) found that a weight gain greater than 3.5 kg/m2 was associated with a 2.7-fold (CI, 2.3- to 3.2-fold) increased risk for developing new reflux symptoms. On the other hand, although weight loss is often recommended as a therapeutic measure in reflux disease (57), the studies yielded conflicting results on its efficacy (58, 59). Small nonrandomized studies suggest that weight loss after bariatric surgery for morbid obesity is associated with an improvement in GERD symptoms (60, 61). Finally, no data suggest that weight loss can affect risk for esophageal adenocarcinoma.
The mechanism of obesity-related esophageal adenocarcinoma is unclear. Of interest, no studies satisfying our inclusion and exclusion criteria assessed the association between obesity and Barrett esophagus. Excluded because of the absence of a control group, Caygill and colleagues' study (62) reported a BMI greater than 30 kg/m2 in 24% of patients in their case series of 102 patients with Barrett esophagus, as compared with an obesity prevalence of 13% in the general population of England. Excluded because of few cases, a case–control study by Chak and colleagues (63) and a cross-sectional study by Gerson and colleagues (64) both reported no apparent increase in risk for obesity in 35 and 27 patients with Barrett esophagus, respectively, although Chak and colleagues reported that the duration of obesity was statistically significantly longer in patients with Barrett esophagus.
Our review suggests that overweight and obesity are risk factors for acid-related esophageal disease. Future studies should examine the mechanism by which obesity causes these complications, as well as the potential effects of weight loss. In the meantime, however, it is prudent to counsel all overweight patients who present with GERD-related diseases that weight loss may help improve symptoms.
Article and Author Information
-
Acknowledgments: The authors thank Loralee Capistrano, MPH, and Jennifer R. Kramer, PhD, for their help in using STATA.
-
Grant Support: Dr. El-Serag is a recipient of the Department of Veterans Affairs Health Services Research and Development Advanced Career Development Award (RCD00-013-2).
-
Potential Financial Conflicts of Interest: None disclosed.
-
Requests for Single Reprints: Hashem B. El-Serag, MD, MPH, Michael E. DeBakey Veterans Affairs Medical Center, 2002 Holcombe Boulevard (152), Houston, TX 77030; e-mail, hasheme{at}bcm.tmc.edu.
-
Current Author Addresses: Drs. Hampel, Abraham, and El-Serag: Michael E. DeBakey Veterans Affairs Medical Center, 2002 Holcombe Boulevard (152), Houston, TX 77030.
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
Related Clinical Content
