Associations among binge eating behavior patterns and gastrointestinal symptoms: a population-based study

Abstract

Background

The psychological symptoms associated with binge eating disorder (BED) have been well documented. However, the physical symptoms associated with BED have not been explored. Gastrointestinal (GI) symptoms such as heartburn and diarrhea are more prevalent in obese adults, but the associations remain unexplained. Patients with bulimia have increased gastric capacity. The objective of the study was to examine if the severity of binge eating episodes would be associated with upper and lower GI symptoms.

Methods

Population-based survey of community residents through a mailed questionnaire measuring GI symptoms, frequency of binge eating episodes and physical activity level. The association of GI symptoms with frequency of binge eating episodes was assessed using logistic regression models adjusting for age, gender, body mass index (BMI) and physical activity level.

Results

In 4096 subjects, BED was present in 6.1%. After adjusting for BMI, age, gender, race, diabetes mellitus, socioeconomic status and physical activity level, BED was independently associated with the following upper GI symptoms: acid regurgitation (P < 0.001), heartburn (P < 0.001), dysphagia (P < 0.001), bloating (P < 0.001) and upper abdominal pain (P < 0.001). BED was also associated with the following lower GI symptoms: diarrhea (P < 0.001), urgency (P < 0.001), constipation (P < 0.01) and feeling of anal blockage (P = 0.001).

Conclusion

BED appears to be associated with the experience of both upper and lower GI symptoms in the general population, independent of the level of obesity. The relationship between increased GI symptoms and physiological responses to increased volume and calorie loads, nutritional selections and rapidity of food ingestion in individuals with BED deserves further study.

Introduction

Gastrointestinal (GI) disorders, especially functional GI disorders, have been shown to induce high health-care utilization and negative impact on quality of life.¹ Chronic GI symptoms such as heartburn and diarrhea are more prevalent in obese adults, but the associations remain unexplained.² Binge eating disorder (BED) occurs in a subset with obesity; rapid food ingestion might lead to GI symptoms but this has been little explored.³ The objective of this study was to examine if binge eating is associated with upper and lower GI symptoms, and to explore the relationships between BED, obesity and GI symptoms.

Obesity has become a major epidemic worldwide. According to the National Center for Health Statistics, more than half of the US population aged 20 or older is overweight (body mass index, BMI): 25.0–29.9 kg/m²), and a quarter are obese (BMI ≥30 kg/m²).⁴ Researchers in Europe have found an association between a higher BMI and the presence of symptoms compatible with gastroesophageal reflux disease.⁵ A population-based case–control study in Norway observed that obesity was associated with a higher prevalence of endoscopic esophagitis.⁶ In a survey of residents of Olmsted County, MN, USA, Delgado-Aros et al.² found that individuals with a higher BMI were more likely to report symptoms attributable to the upper and lower GI tract, including vomiting, upper abdominal pain, bloating and diarrhea. In addition, increases in body weight of over 4.5 kg over a 10-year period are associated with the onset of new GI symptoms.⁷ However, an explanation for this link between obesity and GI symptoms remains to be provided.

Obese individuals with BED may help account for this association. Features of BED include eating a large amount of food in a short period of time, eating rapidly and experiencing a perceived loss of control over eating.^8,9 From a physiological perspective, excessive intake of food over a relatively short time could potentially overcome the functional accommodation and emptying reserve of the stomach, and contribute to the genesis of GI symptoms in obese individuals. BED has been reported to be relatively frequent in obese adults although the exact prevalence is unclear.^10–12 A survey of 1632 overweight individuals in the Midwest of the United States found that approximately 9% of men and 13% of women were likely to be binge eaters.¹³ The psychological factors associated with BED are well documented. Obese binge eaters report higher levels of depression, more negative body image, lower quality of life¹⁴ and lower self-confidence compared to obese non-binge eaters.^15,16 In contrast, the physiological characteristics associated with BED are not well understood.

The effect of specific patterns of eating behavior such as binge eating on the development of both upper and lower GI symptoms in obesity is incompletely defined. Crowell et al.³ found associations between binge eating, obesity and symptoms of irritable bowel syndrome in a small, tertiary referral patient study, but no population-based data are currently available. In addition to perhaps helping design more efficacious interventions for BED, if patterns of food ingestion contribute to the development of unexplained GI symptoms, then attention to eating patterns may provide for a simple, safe and potentially effective nonpharmacological or behavioral method to treat symptoms suggestive of functional GI disorders. It is also conceivable that some individuals presenting for medical treatment for GI symptoms could be better managed by identification and receiving treatment for concurrent BED. This is an important issue given the recent attribution of cardiovascular events, and ischemic colitis to serotonergic agents that were used for the treatment of such GI symptoms.¹⁷ Such vascular risks may be even more pertinent in obese individuals, who often have vascular comorbidity.

In searching for physiological differences in BED, Geliebter and colleagues have demonstrated that the stomach capacity of patients with BED is greater than that of controls. The stomach capacity of non-BED obese individuals,^18,19 measured either with intubated²⁰ or noninvasive imaging, was not significantly different from that of controls.^21,22 We have also observed that increased gastric capacity is associated with lower postprandial symptoms and satiation.²³ Thus, it is thought that due to increased gastric accommodation in BED, there may be less of an upper GI symptom burden, but the increased intake may precipitate colonic discomfort because of an increased osmotic load downstream.

Given the increased gastric capacity of patients with BED, the hypothesis of this study was that BED would be associated with reduced upper abdominal symptoms, but greater lower GI symptoms. The aim of this study was to assess the association between binge eating episodes and upper and lower GI symptoms in a cross-sectional population study. We also assessed the impact of physical activity levels, and other demographic and lifestyle covariables (age, race, educational level, presence of diabetes mellitus, cigarette smoking and alcohol consumption) on the association of binge eating episodes with GI symptoms. The effects of physical activity level on GI symptoms were explored because epidemiological and clinical studies have reported contrasting results on the effects of physical activity level on GI morbidity and function,²⁴ and because it is unknown whether physical activity level moderates any effect of BED on induction of GI symptoms.

Methods

Sampling frame

The responses utilized in this study were obtained from two distinct sampling frames. The first was based on an enumeration of the local community through the medical records linkage system (the Rochester Epidemiology Project, REP) and the second was a purchased list-based sample of noninstitutionalized residents of Olmsted County, MN, who were 18 years and older. This study was approved by Mayo Clinic’s Institutional Research Review Board, and participants were informed regarding the research purpose of the survey. Because Mayo Clinic is considered a covered entity under the Health Insurance Portability and Accountability Act of 1996 (HIPAA) regulations, the Mayo IRB typically requires a signed HIPAA Authorization Form (HAF) for all studies that involve receipt of personal health information, including survey-based investigations. Consistent with the HIPAA guidelines (http://www.hhs.gov/ocr/hipaa), the HAF contained the following information: (1) purpose (title of the study); (2) what patient information is being used or disclosed; (3) who is authorized to receive and use the patient information; (4) who is authorized to disclose patient information; (5) right to refuse to sign and revoke authorization at any time; (6) potential for re-disclosure of patient information; (7) expiration date (for example, ‘end of study’ vs ‘never’) and (8) signature. Information only from those who completed and returned a signed HAF was used in the present analyses.

First sampling frame

The initial sampling frame was developed using the infrastructure of the REP, which provides what is essentially an almost complete ( > 95%) enumeration of the population of Olmsted County, MN.²⁵ This enumeration stems from the medical records linkage system of the Mayo Clinic, which is the major health-care provider in the county.²⁵ The potential of this data system for use in population-based studies has been described previously.^25,26 Using this enumeration of the county as the sampling frame, a gender and age-stratified (10-year age groups, age range 18–65 years) random sample of 4805 county residents was selected. The ethnic constitution of southeastern Minnesota communities is approximately 90% Caucasian, 5% Southeast Asian and 5% other minorities.²⁵ Subjects were randomly selected using the enumeration of the local community provided by the REP and mailed a 29-item survey containing questions about GI symptoms, current height and weight, eating habits, physical activity, education level, race, diabetes mellitus, alcohol and cigarette use and family history of abdominal problems. Two separate mailings were sent in November 2003 and January 2004. Institutionalized subjects, or those who could not be located by mail, subjects who had moved from the Olmsted County from the last contact with the Mayo Clinic and subjects who had denied authorization for research were considered ineligible and excluded. The Mayo Clinic Survey Research Center performed data retrieval and built the database.

Second sampling frame

The second sampling frame utilized a list-appended random digit dial (RDD) sample whereby postal addresses were appended to the RDD telephone numbers if they were found in listed directories. This sample was purchased from an outside survey sampling vendor and included a total of 7000 noninstitutionalized adults aged 18 and older residing in Olmsted County. The questionnaire sent in this mailing was a somewhat expanded version of the initial survey. This second survey included items that assessed mental health status. On this occasion, the data collection protocol consisted of the following steps: an initial mailed survey with a cover letter message explaining the study; a second mailed survey 3 weeks after the initial mailing, again with cover letter message, to nonrespondents to the previous mailing; and nonrespondents to the previous two mailings were sent to the telephone interviewing staff approximately 5 weeks after the initial mailing.

The flow of the surveys is shown in Figure 1. A total of 943 subjects responded to the initial survey (response rate 26%) and 3181 subjects (46% of the sample) responded to the repeat survey. Responders and nonresponders were similar sociodemographically (see results).

Figure 1.

Summary of survey recruitment.

Questionnaire, binge eating patterns and exercise definitions GI symptoms

All the items had been previously individually validated. The questions on GI symptoms were based on the Bowel Disease Questionnaire but used a five-point Likert scale.²⁷ Constipation was defined as the report of less than three bowel movements per week. Diarrhea was defined as the report of more than three bowel movements per day. An overall symptom score (mean over all 16 symptom items) was also computed using the five-point Likert scale values for each item.

Binge eating

Key questions on eating behavior were based on the Questionnaire on Eating and Weight Patterns-Revised and have been shown to reliably predict BED in previous studies.¹³ The subjects were asked:

1. In the last three months, did you often eat within any two-hour period what most people would regard as an unusually large amount of food? (choice Yes/No),

2. If yes, during the times when you ate this way, did you often feel you could not stop eating or control what or how much you were eating? (choice Yes/No),

3. If yes, over the past three months, how often did you feel you could not stop eating or control what or how much you were eating? (choices from 1 to 7 days per week).

Subjects were classified as ‘over-eaters’ if they provided a positive response to question 1. Subjects responding positively to the first two questions and providing a frequency of at most two episodes per week were characterized as ‘binge eaters’ as many would not fully meet diagnostic criteria for BED. We applied this more conservative approach due to use of a mailed survey and the lack of clinical confirmation. Those with three or more days per week of binge eating episodes were classified as having BED, as they met diagnostic criteria for this eating behavior pattern and reported having a greater frequency of binge eating episodes than the rest of the survey sample.¹³

Physical activity level

Physical activity was measured as follows:²⁸

1. Aside from any work you do at home or at a job, do you do anything regularly, that is, on a daily basis that helps you keep physically fit? (choice Yes/No),

2. How often in your free time, do you take part in moderate physical activity (such as bowling, golf, light sports or physical exercise, gardening, taking long walks)? (choices more than 4 times a week, 2–4 times a week, about once a week, a few times a month, a few times a year, rarely or never),

3. How often, in your free time, do you take part in vigorous physical activity (such as jogging, racket sports, swimming, aerobics, strenuous sports)? (frequency choices as in previous question).

The two questions eliciting self-report of exercise levels:²⁸ ‘moderate’(for example, walking, gardening) and ‘vigorous’ (for example, swimming, jogging, aerobics), both on six-point Likert frequency scales (rarely to daily) were used to compute a quantitative physical activity level score. A greater score was indicative of higher physical activity levels.

For the purpose of summarizing symptom data, in addition to the physical activity score, the responses to these questions were also used to categorize the participants into different levels of physical activity:

1. High exercisers, reporting engaging in vigorous exercise at least a few times a week.

2. Moderate exercisers, reporting engaging in moderate exercise at least two times a week and vigorous exercise at most a few times per month.

3. Light exercisers, reporting engaging in moderate exercise at least a few times a month to once a week and vigorous exercise at most a few times per month.

4. Rare exercisers, reporting vigorous or moderate exercise at most a few times a year.

Body mass index

The BMI was calculated from self-reported height and weight. Data from the Third National Health and Nutrition Examination Survey suggest that the misclassification of overweight and obese subjects obtained using self-report BMI is minimal.²⁹ BMI levels were based on the WHO classification of obesity.³⁰

Mental health status

Questions from the vitality and mental health scales were used from the SF-12.³¹ The subjects were asked the following questions with the stem of ‘How much of the time during the past 4 weeks’:

1. Did you have a lot of energy?

2. Have you felt downhearted and blue?

The mean of these two responses (six-point Likert scale: 1 = ‘all of the time’ to 6 = ‘none of the time’), first reversing the scale for the second question, was computed for each subject. In addition, a question from the SF-12 on whether physical health or emotional problems interfered with social activities during the past 4 weeks was included as a covariate along with the mean response of the two questions above to adjust for mental health and physical/emotional status.

Demographic and other risk factors

Alcohol consumption and smoking status were asked using single items covering current use (past 30 days). Other questions also asked about any history of abdominal pain in first-degree relatives, any diagnosis of diabetes mellitus, education level attained and ethnicity.

Statistical analysis

For the main analyses, respondents from both sampling frames—the enumeration of the local community through the REP medical records linkage system and the purchased list-based sample of Olmsted County residents—were combined. The associations between binge eating episodes, physical activity levels and individual upper and lower GI symptoms were assessed using multiple logistic regression, a generalized logit link function. The estimated odds ratios (OR) and 95% confidence intervals (CI) for categories of symptom frequency were estimated from the coefficients (and their standard errors) in the logistic regression models. The OR for each GI symptom (separately) in binge eaters and those with BED (both relative to no binge eating) were estimated adjusting for age, gender, BMI, physical activity level score and version of the survey in which the responses were obtained. Additional models to assess the potential ‘interaction’ of binge eating episodes and BMI were also examined by constructing dummy regression variables corresponding to combinations of binge eating (none/any) and BMI category (underweight, normal weight, overweight and obese) with normal weight no binge eating as the reference level. A test for interaction effects was based on the difference in log likelihoods for the ‘no interaction’ and ‘interaction’ models. Furthermore, separate models in subjects with BMI less than 25 and in subjects with BMI equal to or greater than 25 were also summarized.

Separate logistic models using just subjects responding to the version of the survey containing the SF-12 were also examined adjusting for ‘mental health status’ (two questions characterizing mental health status and the question regarding physical/emotional problems impact on social activities). The univariate association of the three SF-12 measures with binge eating episodes was assessed using contingency table analyses (χ²-test) and their univariate association with overall GI symptom score assessed using Spearman’s correlations. A multiple linear regression model using binge eating categories (binge eating, BED and no binge eating as the reference level) to predict the overall GI symptom score was also examined, including age, gender, BMI, exercise score and the SF-12 measures as covariates.

An additional logistic model to predict binge eating (none, binge eating and BED) was examined with age, gender, BMI and physical activity score as the predictor variables.

Results

Prevalence and predictors of binge eating

The characteristics of all responders are shown in Table 1 (by binge eating) and Table 2 (by BMI category). Overall, 15% of responders (615 of 4096 providing data on binge eating episodes) reported experiencing some episodes of overeating within the previous 3 months, whereas 250 (6.1%) reported experiencing binge eating and 111 (2.7%) of these reported having BED (Table 1).

Table 1.

Age, BMI, exercise score and GI symptom score by binge eating categories

	Overall			Missinga			No binge eating			Binge eating			Binge eating disorder
	N	Mean (s.e.)	Median (range)	N	Mean (s.e.)	Median (range)	N	Mean (s.e.)	Median (range)	N	Mean (s.e.)	Median (range)	N	Mean (s.e.)	Median (range)
Age	4124	51.0 (0.2)	51.0 (18.0–100.0)	28	56.4 (3.7)	59.5 (19.1–89.0)	3481	52.1 (0.3)	52.0 (18.0–100.0)	365	44.4 (0.7)	44.0 (18.3–86.0)	250	45.1 (0.8)	45.0 (18.0–84.0)
BMI	4124	27.3 (0.1)	26.5 (3.2–69.1)	28	24.7 (0.7)	24.2 (19.5–33.1)	3481	26.9 (0.1)	26.1 (3.2–69.1)	365	29.2 (0.3)	28.3 (16.9–59.6)	250	31.1 (0.5)	30.5 (16.5–52.3)
Exercise	4119	9.9 (0.1)	9.0 (2.0–18.0)	23	11.6 (0.8)	11.0 (4.0–18.0)	3481	9.9 (0.1)	9.0 (2.0–18.0)	365	10.5 (0.2)	10.0 (3.0–18.0)	250	9.5 (0.3)	8.0 (3.0–18.0)
score
Symptom	4104	1.5 (0.0)	1.4 (1.0–4.1)	28	1.8 (0.1)	1.9 (1.1–3.4)	3462	1.5 (0.0)	1.4 (1.0–4.1)	365	1.6 (0.0)	1.5 (1.0–3.2)	249	1.9 (0.0)	1.8 (1.0–3.9)
score

Abbreviation: BMI, body mass index.

^aMissing data on binge eating behavior.

Table 2.

Age, exercise and GI symptom scores by BMI categories

	Underweight ( < 18.51)			Normal (18.5–25)			Overweight (25–30)			Obese ( ≥ 30)
	N	Mean (s.e.)	Median (range)	N	Mean (s.e.)	Median (range)	N	Mean (s.e.)	Median (range)	N	Mean (s.e.)	Median (range)
Age	45	48.7 (3.2)	44.0 (18.2–96.0)	1488	48.4 (0.4)	48.0 (18.0–99.0)	1541	52.6 (0.4)	52.0 (18.0–94.0)	1050	52.4 (0.4)	52.2 (18.2–100.0)
Exercise score	45	9.8 (0.6)	10.0 (3.0–18.0)	1484	10.8 (0.1)	10.0 (2.0–18.0)	1540	10.0 (0.1)	9.0 (2.0–18.0)	1050	8.5 (0.1)	7.0 (3.0–18.0)
Symptom score	45	1.5 (0.1)	1.4 (1.0–3.4)	1478	1.5 (0.0)	1.4 (1.0–4.1)	1536	1.5 (0.0)	1.4 (1.0–3.8)	1045	1.6 (0.0)	1.5 (1.0–4.1)

The factors predicting binge eating are summarized in Table 3. Increasing age reduced the odds for overeating and binge eating (all categories), and whereas males had increased odds for binge eating, they had decreased odds (relative to females) for BED. Greater BMI values were associated with increased odds for all categories of binge eating, whereas physical activity level was only modestly associated with binge eating.

Table 3.

Features associated with binge eating (vs no binge eating)

Predictors	OR (95% CI) for overeaters	OR (95% CI) for binge eaters	OR (95% CI) for binge eating disorder
Age (per 10 years)	0.66 (0.61, 0.71)	0.69 (0.61, 0.79)	0.71 (0.62, 0.83)
Gender
Female	1.0 (ref)	1.0 (ref)	1.0 (ref)
Male	2.2 (1.7, 2.8)	0.9 (0.6, 1.3)	0.5 (0.3, 0.7)
BMI (per unit)	1.09 (1.09, 1.11)	1.12 (1.10, 1.15)	1.13 (1.10, 1.16)
Exercise score (per unit)	1.03 (1.00, 1.06)	1.03 (0.99, 1.07)	0.98 (0.93, 1.03)

Abbreviation: BMI, body mass index. Odds ratios (95% confidence intervals) indicated in bold type are significant at an a level of 0.05.

Association between binge eating and GI symptoms

The symptom distribution (clinically important frequency levels, for example, often or very often) by binge eating categories is summarized in Table 4. The adjusted OR (95% CI) for these clinically important frequency levels of individual upper and lower GI symptoms in binge eaters relative to non-binge eaters are shown in Figure 2 and Figure 3.

Table 4.

Distribution of symptoms by binge eating category

Symptom	No binge eating (N = 3481) N (%)	Binge eating (N = 365) N (%)	Binge eating disorder (N = 250) N (%)
Abdominal pain	239 (6.9)	27 (7.4)	34 (13.6)
Fullness	87 (2.5)	5 (1.4)	6 (2.4)
Food staying in the stomach	128 (3.7)	17 (4.7)	14 (5.6)
Bloating	247 (7.1)	25 (6.8)	41 (16.4)
Acid regurgitation	193 (5.5)	21 (5.8)	36 (14.4)
Heartburn	211 (6.1)	21 (5.8)	35 (14.0)
Nausea	48 (1.4)	7 (1.9)	6 (2.4)
Vomiting	63 (1.8)	7 (1.9)	10 (4.0)
Dysphagia	58 (1.7)	6 (1.6)	11 (4.4)
Stool blockage	100 (2.9)	14 (3.8)	14 (5.6)
Diarrhea	188 (5.4)	31 (8.5)	41 (16.4)
Constipation	98 (2.8)	12 (3.3)	14 (5.6)
Lumpy/hard stools	181 (5.2)	19 (5.2)	20 (8.0)
Loose/watery stools	197 (5.7)	28 (7.7)	33 (13.2)
Fecal urgency	178 (5.1)	22 (6.0)	40 (16.0)
Fecal incontinence	131 (3.8)	13 (3.6)	23 (9.2)

Figure 2.

(a, b) Odds ratios (95% confidence intervals) for clinically important upper gastrointestinal symptom reporting in overeating (O), and separately (B) binge eating, each relative to ‘no binge eating’ (adjusted for age, gender, body mass index (BMI), physical activity score and version of the survey in which the responses were obtained). Odds ratios with 95% CI that do not contain the value 1.0 (horizontal reference line) are significant at an α level of 0.05. Reg, regurgitation.

Figure 3.

(a, b) Odds ratios (95% confidence intervals) for clinically important lower gastrointestinal symptom reporting in overeating (O), and separately (B) binge eating, each relative to ‘no binge eating’ (adjusted for age, gender, body mass index (BMI), physical activity score and version of the survey in which the responses were obtained). Odds ratios with 95% CI that do not contain the value 1.0 (horizontal reference line) are significant at an α level of 0.05.

BED was associated with the following upper GI symptoms: heartburn (OR 2.2, 95% CI 1.4–3.3, P < 0.001), acid regurgitation (OR 2.3, 95% CI 1.5–3.6, P < 0.001), dysphagia (OR 3.3, 95% CI 1.6–6.8, P < 0.001), bloating (OR 3.6, 95% CI 2.3–5.5, P < 0.001) and upper abdominal pain (OR 2.3, 95% CI 1.4–3.6, P < 0.001). BED was associated with the following lower GI symptoms: diarrhea (OR 3.1, 95% CI 2.0–4.6, P < 0.001), urgency (OR 3.9, 95% CI 2.6–6.1, P < 0.001), constipation (OR 2.2, 95% CI 1.2–4.1, P < 0.01) and a feeling of anal blockage (OR 2.7, 95% CI 1.5–5.0, P = 0.001). All associations were adjusted for age, gender, BMI (as a continuous covariate), physical activity level and version of the survey.

Obesity and GI symptoms

The adjusted OR for the clinically important frequency levels of symptoms in overweight and, separately, obese categories (both relative to underweight/normal) are given in Table 5. For example, obesity was associated with increased odds for reporting of acid regurgitation (OR 3.39, 95% CI 2.36–4.87, P < 0.001), heartburn (OR 3.11, 95% CI 2.20–4.39, P < 0.001), diarrhea (OR 1.64, 95% CI 1.16–2.32, P = 0.005), loose stools (OR 1.63, 95% CI 1.11–2.29, P = 0.005) and urgency (OR 1.46, 95% CI 1.03–2.09, P = 0.036) after adjusting for age and gender, binge eating, physical activity level and version of the survey. Table 6 shows the OR estimated separately for under/normal weight groups and overweight/obese groups. The tests for interaction indicated significantly increased odds for vomiting in binge eating (relative to no binge eating) in normal weight, but not an increased odds in overweight or in obese subjects (P = 0.005, difference in model log likelihoods). A similar differential pattern was also detected for lumpy/hard stools (P = 0.025, difference in model log likelihoods) with an increased odds for lumpy/hard stools in normal weight subjects reporting binge eating, but not in overweight subjects with and without binge eating.

Table 5.

Odds for symptom reporting in overweight and obese (relative to normal/underweight)

Symptom	Body mass index category
	Overweight		Obese
	ORa	95% CI	ORa	95% CI
Abdominal pain	0.90	(0.66, 1.23)	1.29	(0.93, 1.78)
Fullness	0.90	(0.55, 1.46)	0.89	(0.52, 1.53)
Food staying in the stomach	1.37	(0.91, 2.07)	1.76	(1.15, 2.70)
Bloating	0.98	(0.73, 1.33)	1.07	(0.77, 1.48)
Acid regurgitation	2.00	(1.39, 2.86)	3.39	(2.36, 4.87)
Heartburn	1.64	(1.16, 2.31)	3.11	(2.20, 4.39)
Nausea	0.70	(0.35, 1.40)	1.46	(0.77, 2.75)
Vomiting	1.05	(0.59, 1.87)	1.70	(0.96, 3.02)
Dysphagia	0.53	(0.30, 0.95)	0.66	(0.36, 1.18)
Anal blockage	0.63	(0.41, 0.99)	0.74	(0.46, 1.17)
Diarrhea	1.35	(0.97, 1.88)	1.64	(1.16, 2.32)
Constipation	0.72	(0.47, 1.11)	0.50	(0.30, 0.85)
Lumpy/hard stools	0.99	(0.72, 1.37)	0.49	(0.33, 0.75)
Loose/watery stools	1.09	(0.78, 1.53)	1.63	(1.15, 2.29)
Fecal urgency	1.09	(0.77, 1.54)	1.46	(1.03, 2.09)
Fecal incontinence	0.98	(0.66, 1.46)	1.36	(0.91, 2.04)

Odds ratios (95% confidence intervals) indicated in bold type are significant at an α level of 0.05.

^aOdds ratio (95% CI) from logistic regression model adjusting for age, gender, binge eating category, physical activity score and version of the survey.

Table 6.

Odds for symptom reporting in eating pattern (overeating vs none, and binge eating vs none) separately by BMI category

Symptom	Eating pattern	Body mass index categories
		Underweight/normal		Overweight/obese
		ORa	95% CI	ORa	95% CI
Abdominal pain	Overeating	1.43	(0.62, 3.29)	1.33	(0.78, 2.28)
	Binge eating	2.98	(1.20, 7.44)	2.26	(1.35, 3.76)
Fullness	Overeating	1.16	(0.33, 4.05)	0.33	(0.08, 1.36)
	Binge eating	2.97	(0.84, 10.48)	0.51	(0.15, 1.69)
Food staying in the stomach	Overeating	2.48	(0.90, 6.88)	1.24	(0.65, 2.35)
	Binge eating	2.85	(0.93, 8.76)	1.42	(0.70, 2.86)
Bloating	Overeating	1.56	(0.65, 3.72)	1.38	(0.80, 2.39)
	Binge eating	6.54	(2.62, 16.32)	3.07	(1.88, 4.99)
Acid regurgitation	Overeating	0.80	(0.23, 2.74)	1.33	(0.77, 2.29)
	Binge eating	3.86	(1.39, 10.73)	2.48	(1.55, 3.96)
Heartburn	Overeating	0.55	(0.16, 1.87)	1.09	(0.63, 1.86)
	Binge eating	4.83	(1.94, 12.01)	2.22	(1.37, 3.61)
Nausea	Overeating	2.88	(0.90, 9.27)	0.76	(0.22, 2.58)
	Binge eating	4.92	(1.35, 17.96)	0.66	(0.19, 2.26)
Vomiting	Overeating	2.44	(0.67, 8.93)	0.65	(0.23, 1.85)
	Binge eating	6.86	(2.37, 19.87)	0.86	(0.33, 2.25)
Dysphagia	Overeating	0.56	(0.07, 4.34)	1.30	(0.49, 3.42)
	Binge eating	12.68	(4.47, 35.92)	1.53	(0.57, 4.09)
Stool blockage	Overeating	2.74	(1.09, 6.92)	1.87	(0.86, 4.10)
	Binge eating	3.60	(1.30, 9.95)	2.63	(1.23, 5.64)
Diarrhea	Overeating	1.50	(0.56, 4.01)	1.73	(1.08, 2.76)
	Binge eating	8.29	(3.67, 18.71)	2.56	(1.62, 4.03)
Constipation	Overeating	1.67	(0.62, 4.50)	1.33	(0.58, 3.04)
	Binge eating	2.69	(0.99, 7.34)	1.92	(0.90, 4.08)
Lumpy/hard stools	Overeating	1.89	(0.80, 4.49)	1.42	(0.75, 2.70)
	Binge eating	3.27	(1.41, 7.57)	1.48	(0.75, 2.92)
Loose/watery stools	Overeating	2.20	(0.91, 5.31)	1.75	(1.04, 2.96)
	Binge eating	5.93	(2.51, 13.97)	1.97	(1.18, 3.30)
Fecal urgency	Overeating	1.14	(0.38, 3.37)	1.94	(1.11, 3.39)
	Binge eating	4.33	(1.73, 10.82)	4.21	(2.59, 6.83)
Fecal incontinence	Overeating	0.52	(0.07, 3.94)	1.65	(0.87, 3.13)
	Binge eating	5.13	(1.99, 13.25)	2.69	(1.51, 4.79)

Odds rations (95% confidence intervals) indicated in bold type are significant at an α level of 0.05.

^aOdds ratios for the designated category of the symptom in overeating vs no binge eating; and separately, binge eating vs no binge eating, from logistic regression models adjusting for age, gender, physical activity score and version of the survey.

Mental health status and GI symptoms

Odds for symptom reporting in binge eaters and separately, those with BED (each relative to no binge eating) were estimated adjusting for mental health status and physical/emotional impact on social activities. Table 7 shows the results adjusting and not adjusting for these SF-12 questions. It should be noted that the unadjusted OR were computed in the subset of subjects that were mailed the second version of the questionnaire because only the second survey contained these items. Adjusting for these covariates typically decreased the odds for reporting GI symptoms, suggesting that mental and emotional health status may be a confounder of binge eating effects. Indeed, significant associations (P < 0.001) between binge eating and the SF-12 questions used to define mental health status were observed, and significant (P < 0.001) univariate associations between these measures and the overall GI symptom score. The multiple linear regression model for the overall GI symptom score did however indicate binge eating categories to be significant (P < 0.05) independent predictors of symptom score.

Table 7.

Odds for symptom reporting for eating pattern (overeating vs none, and binge eating vs none) unadjusted, and adjusted for mental health status (using questions from SF-12)

Symptom	Eating pattern	Unadjusted a		Adjusted
		For mental health status
		ORa	95% CI	ORa	95% CI
Abdominal pain	Overeaters	1.20	(0.69, 2.08)	1.02	(0.57, 1.80)
	Binge eaters	2.22	(1.29, 3.81)	1.21	(0.69, 2.13)
Fullness	Overeaters	0.33	(0.06, 1.08)	0.25	(0.06, 1.08)
	Binge eaters	0.95	(0.36, 2.51)	0.46	(0.17, 1.24)
Food staying in the stomach	Overeating	1.14	(0.57, 2.31)	0.97	(0.47, 2.02)
	Binge eating	1.10	(0.48, 2.53)	0.63	(0.27, 1.48)
Bloating	Overeating	1.82	(1.07, 3.11)	1.62	(0.92, 2.83)
	Binge eating	2.87	(1.66, 4.96)	1.63	(0.92, 2.90)
Acid regurgitation	Overeating	1.12	(0.62, 2.01)	0.99	(0.54, 1.80)
	Binge eating	2.21	(1.32, 3.73)	1.45	(0.85, 2.47)
Heartburn	Overeating	1.07	(0.61, 1.86)	0.94	(0.53, 1.66)
	Binge eating	1.89	(1.11, 3.21)	1.30	(0.76, 2.23)
Nausea	Overeating	0.70	(0.21, 2.37)	0.55	(0.16, 1.91)
	Binge eating	1.58	(0.62, 4.02)	0.75	(0.29, 1.95)
Vomiting	Overeating	0.52	(0.16, 1.73)	0.46	(0.14, 1.54)
	Binge eating	2.13	(1.01, 4.52)	1.40	(0.65, 3.00)
Dysphagia	Overeating	1.42	(0.58, 3.45)	1.31	(0.54, 3.22)
	Binge eating	2.71	(1.15, 6.39)	1.91	(0.80, 4.55)
Stool blockage	Overeating	1.49	(0.69, 3.22)	1.35	(0.62, 2.96)
	Binge eating	2.05	(0.93, 4.53)	1.27	(0.57, 2.85)
Diarrhea	Overeating	1.67	(1.02, 2.72)	1.56	(0.96, 2.56)
	Binge eating	3.36	(2.11, 5.36)	2.73	(1.70, 4.40)
Constipation	Overeating	1.14	(0.51, 2.55)	1.03	(0.46, 2.34)
	Binge eating	2.56	(1.30, 5.01)	1.79	(0.90, 3.56)
Lumpy/hard stools	Overeating	1.30	(0.69, 2.45)	1.18	(0.62, 2.24)
	Binge eating	2.03	(1.07, 3.87)	1.42	(0.74, 2.74)
Loose/watery stools	Overeating	2.22	(1.32, 3.72)	2.00	(1.18, 3.38)
	Binge eating	2.79	(1.68, 4.71)	1.91	(1.12, 3.26)
Fecal urgency	Overeating	1.63	(0.89, 3.00)	1.50	(0.81, 2.78)
	Binge eating	4.07	(2.43, 6.83)	2.62	(1.53, 4.46)
Fecal incontinence	Overeating	1.17	(0.55, 2.50)	1.11	(0.52, 2.38)
	Binge eating	3.57	(1.98, 6.45)	2.83	(1.55, 5.17)

Odds ratios (95% confidence intervals) indicated in bold type are significant at an α level of 0.05.

^aNote that the unadjusted values use only the subjects who responded to the version of the survey containing the SF-12.

Differences in populations and effect of individual surveys

Demographics (including age, gender and proportion in single BMI classes) were not significantly associated with whether respondents participated in the first or the second survey. It is very unlikely that subjects participated in both surveys. Whereas there is no real mechanism to confirm this due to the de-identification process, it is unlikely that respondents would complete two copies of the same survey sent to them in a short period of time. The association between binge eating and GI symptoms was similar on both surveys based on additional logistic regression models that contained interaction terms for survey version by binge eating categories. These terms were not significant and thus dropped from the final models. The final models did retain a term for survey version, which typically indicated smaller odds for symptom reporting on the second version of the survey (relative to the first survey). Finally, the overall sample of responders was similar in terms of age and gender distribution to the 2000 Olmsted County population demographic characteristics.

Discussion

This study reports, in a population-based sample, the associations between frequency of binge eating episodes, physical activity level and individual GI symptoms. In 4096 subjects from the Olmsted County population, BED was associated with several upper and lower GI symptoms, independent of BMI status or level of physical activity.

Previous population-based studies have suggested that obesity is associated with increased prevalence of gastroesophageal reflux (both symptomatic and endoscopically identified esophagitis).^5,6 Obesity has also been linked to other upper and lower GI symptoms, including bloating, upper abdominal pain, diarrhea (but not constipation) and vomiting.² In this study, the prevalence of GI symptoms and the associations of obesity with GI symptoms were very consistent with previous population-based studies, further supporting our view that the present population sample is comparable with past random samples drawn from Olmsted County where response rates were high.²

However, this study addresses the relative paucity of information on the potential relationship between BED and the pathogenesis of GI symptoms in people in the community. Crowell et al.,³ in the only case–control study in a tertiary referral sample of obese outpatients, reported associations between specific GI symptoms and binge eating, although only 119 obese patients and 19 controls were evaluated. These authors found significantly more frequent reports of nausea, vomiting, bloating, abdominal pain and dyschezia (sense of anal blockage or difficulty with evacuation) in obese binge eaters compared to controls.³ We could not confirm an association of binge eating with nausea and vomiting; however, in addition to the expected gastroesophageal reflux symptoms, we found associations of BED with bloating, abdominal pain, constipation, a sense of anal blockage and diarrhea in this general population sample.

The pathophysiological bases underlying the higher prevalence of upper GI symptoms in binge eaters are unclear. This higher prevalence was also contrary to the study expectations. Previous research has demonstrated that BMI and stomach volume are independent factors in determining postprandial satiation.³² BED is also associated with increased gastric volume/capacity,^18,20 which should be protective against increased upper GI symptoms in response to food ingestion. Conceivably, the delivery of unusually large amounts of food to the stomach may exceed the large gastric capacity in binge eaters, and it may be inadequately counteracted by gastric reflex adaptive relaxation,³³ causing higher wall tension³⁴ and resulting in the perception of abdominal pain and bloating. Studies using single-photon emission computed tomography imaging show fasting gastric volumes predict food intake in overweight and obese individuals,³² as well as the postprandial symptoms of fullness and bloating in response to a challenge meal.²³ However, Geliebter and colleagues^18,20 have shown that gastric capacity in obese subjects with bulimia is significantly greater than non-bulimic control individuals, irrespective of body weight. The increased gastric capacity may represent an adaptive mechanism to excessive food intake.¹⁹ Conversely, lower fasting volumes are associated with the presence of post-meal symptoms in functional dyspepsia.³⁵ The rapid pattern of ingestion of large amounts of food appears to be important given the normal stomach capacity of obese people with BMIs of up to 45.²¹ In binge eaters, the large amounts of food eaten appear to counteract the mechanisms of lower esophageal sphincter competence,³⁶ possibly resulting in larger quantities of gastric refluxate reaching the distal esophagus and contributing to acid regurgitation and heart-burn. In turn, acid reflux could be the underlying abnormality in subjects complaining of dysphagia.³⁷

The potential mechanisms for generation of lower GI symptoms in binge eaters also need exploration. One potential hypothesis is that larger and more rapidly delivered food boluses to the small intestine represent larger osmotic loads, resulting in intestinal secretion, and a prominent colonic motor response resulting in increased delivery of stool to the distal colon, increasing stool volumes, reduction in stool consistency, leading to diarrhea and urgency. Whereas this study did not inquire about eating more rapidly than usual, one of the important clinical features of BED is eating more rapidly than usual. It is also possible that abnormally high postprandial secretion of gut peptides or neurotransmitters (such as cholecystokinin or serotonin) occurs in the presence of meals that are much larger than usual, contributing to alterations in bowel sensorimotor function.^38,39 These data suggest that detailed GI and colonic transit, hormonal responses to meal ingestion and colonic motor function should be investigated further in patients with bulimia, and particularly in those with BED. Irrespective of the pathophysiology and mechanisms involved, it is intriguing that the association of higher BMI alone with symptoms was relatively modest after adjusting for demographic factors and for binge eating behavior, suggesting eating patterns are more closely linked, than bodily habitus, to symptom generation in the GI tract.

Whereas many of the GI symptoms in patients with eating disorders may be attributable to the eating pattern and the volume or calorie content of food ingested, our hypothesis is that the associations with constipation and sense of anal blockage are independent of feeding issues, and are more likely to result from associated pelvic floor or anal sphincter dyssynergia. We have previously noted the association of another eating disorder, rumination syndrome, in tertiary referral patients attending a program for spastic rectal evacuation disorders.⁴⁰ A retrospective review suggested that increased BMI is not associated with delayed colonic transit; however, it is associated with reduced colonic compliance and pain sensation during distension of an intracolonic balloon.⁴¹ Prospective studies are, therefore, required to assess rectal evacuation function, and colonic sensory and motor physiology in the colon of obese individuals and people with binge eating who may have normal weight.

BED, on the other hand, may be associated with upper and lower GI symptoms because it merely represents a marker of underlying psychological distress rather than reflecting a direct causal relationship. We did observe an association between BED and the mental health score used, and adjusting for this score attenuated all of the associations of BED with GI symptoms except for diarrhea and fecal urgency. Further work is needed using standardized psychiatric measures to determine if anxiety or depression, or somatization, confounds any association of dyspepsia and constipation with BED. It is also possible that the type of foods eaten during a binge episode contributes to GI symptoms because individuals with BED tend to select higher fat foods during a binge episode^42,43 compared to their food selection on non-binge eating days. This potential association between nutrient intake and GI symptoms should be explored in future studies.

This study has other limitations. The low response rate to the initial survey led us to conduct a repeat survey using different methodology that did include direct telephone contact with nonresponders. Although the response rate to the repeat survey was still not ideal, we were able to accrue data for a sample four times larger than the initial one. Moreover, a comparison of our initial and final sample with a recent survey on obesity and GI symptoms conducted in the same population setting revealed very similar distributions of age, gender and body weight categories between the two samples, suggesting demographic homogeneity.² We believe that our sample is representative enough and is also large enough to portray a reliable representation of the population in question. We do not have the means to ascertain the proportion of binge eaters in the individuals not responding to our survey; there is the possibility that subjects affected by BED or by GI symptoms may be somewhat more likely to return our questionnaires, although the rate of binge eating identified is comparable with previous estimates in this part of the United States.¹³ There exists the possibility that in combining the respondents from the two sampling frames utilized in the current study, a certain amount of duplication in response may have impacted the results. Whereas there is no real mechanism to confirm or dispel this possibility due to the de-identification process, we expect that respondents would be unlikely to complete two copies of the same survey sent to them in such a short time. Moreover, the likely impact of a few redundant responses would be minimal. Another limitation pertains to the screening tools used for binge eating behavior. Our study was based on a short self-report questionnaire that, although valid, included only three screening questions for BED. Thus, there is the possibility of having overestimated the prevalence of BED,^44,45 and in future studies a structured clinical interview to assess BED may clarify this issue.⁴⁶

The use of self-reported weight may also have been problematic as studies have shown that adults will under-report their weight.^47,48 A final limitation of this study is that we did not specifically focus on postprandial occurrence of symptoms, but one might expect an even closer relationship of postprandial symptoms with eating behavior and therefore the significant associations likely represent conservative estimates of the true associations between binge eating and GI symptoms.

In conclusion, the impact of BED as a determinant of GI morbidity at a population level is likely to have clinical relevance. However, further insight into the potential mechanisms need to be sought; these epidemiological data provide the basis for planning the studies of GI physiology in eating disorders such as measurements of GI and colonic transit, gastric volume and rectal evacuation dynamics. This study also highlights the potential importance to screen for eating disorders such as BED among patients presenting with GI symptoms because it is conceivable that behavioral interventions (for example, on the rate and quantity of ingested food) may resolve their symptoms without the need to use pharmacological agents.