Gastrointestinal diagnoses in patients with eating disorders: a retrospective cohort study 2010–2020

Abstract

Background and Aims:

Gastrointestinal (GI) disorders are common in patients with eating disorders. However, the temporal relationship between GI and eating disorder symptoms has not been explored. We aimed to evaluate GI disorders among patients with eating disorders, their relative timing, and the relationship between GI diagnoses and eating disorder remission.

Methods:

We conducted a retrospective analysis of patients with an eating disorder diagnosis who had a GI encounter during 2010 to 2020. GI diagnoses and timing of eating disorder onset were abstracted from chart review. Coders applied DSM-5 criteria for eating disorders at the time of GI consult to determine eating disorder remission status.

Results:

Of 344 patients with an eating disorder diagnosis and GI consult, the majority (255/344, 74.2%) were diagnosed with an eating disorder prior to GI consult (preexisting eating disorder). GI diagnoses categorized as functional/motility disorders were most common among the cohort (57.3%), particularly in those with preexisting eating disorders (62.5%). 113 (44.3%) patients with preexisting eating disorders were not in remission at GI consult, which was associated with being underweight (OR 0.13, 95% CI 0.04–0.46, p<0.001) and increasing number of GI diagnoses (OR 0.47 per diagnosis, 95% CI 0.26–0.85, p=0.01).

Conclusions:

Eating disorder symptoms precede GI consult for most patients, particularly in functional/motility disorders. As almost half of eating disorder patients are not in remission at GI consult, GI providers have an important role in screening for eating disorders. Further prospective research is needed to understand the complex relationship between eating disorders and GI symptoms.

Graphical Abstract

Among 344 patients with eating disorder diagnoses and GI visits, eating disorder symptoms precede GI consult for most patients, particularly in those with functional/motility disorders. 44% of eating disorder patients are not in remission at GI consult, and GI providers have an important role in screening for eating disorders.

Introduction

Eating disorders are estimated to have a lifetime prevalence of 14.3% for males and 19.7% for females,¹ and have high rates of hospitalizations and utilization of outpatient care, associated with poor health-related quality of life.² In addition to the well-described relationship between eating disorders and other psychiatric conditions, many patients with eating disorders experience gastrointestinal (GI) symptoms,^3,4 and patients with eating disorders are concentrated disproportionately in GI clinics.^5–7

GI symptoms in eating disorders can span the entire GI tract and may precede, augment or perpetuate disordered eating behavior (see Figure 1). However, the timing of the development of GI symptoms relative to eating disorders has not been well elucidated in the literature, and whether eating disorders induce GI symptoms or conversely, whether GI symptoms lead to eating disorders, is unclear. In a systematic review of 29 studies evaluating neurogastroenterology patients with eating disorders, the causative relationship between GI symptoms and eating disorders could not be determined.⁸

Figure 1.

Possible temporal relationships between disordered eating symptoms and gastrointestinal symptoms. A. Eating disorder symptoms and sequelae may lead to GI symptoms. B. GI symptoms may persist even when eating disorders are in remission.

C. Eating disorders may present with GI symptoms before eating disorder diagnosis.

D. GI symptoms or their management may lead to disordered eating. GI: gastrointestinal

There is likely a bidirectional relationship between GI symptoms and disordered eating, where GI symptoms may contribute to and exacerbate preexisting eating disorder symptoms, and disordered eating may contribute to physiologic changes in the GI tract. Nonetheless, data describing the timing of GI presentations relative to eating disorder onset and the relationship to eating disorder remission is lacking. In this retrospective cohort study, we aimed to evaluate GI disorders among patients with eating disorders, their relative timing, and the relationship between GI diagnoses and eating disorder remission status.

Methods

We assembled a retrospective cohort of patients with a documented diagnosis of an eating disorder (anorexia nervosa (AN), bulimia nervosa (BN), Eating Disorder Not Otherwise Specified (EDNOS), Binge-Eating Disorder (BED), Other Specified Feeding and Eating Disorder (OSFED), Avoidant/Restrictive and Feeding Intake Disorder (ARFID)), and an outpatient GI encounter at Massachusetts General Hospital between 2010–2020. Patients were identified using the Research Patient Data Registry (RPDR)—an electronic database of all patients seeking care within the MassGeneral Brigham healthcare system which is automatically populated with data from multiple sources including clinical encounters, radiology, procedures, laboratory tests, medications, inpatient stays, operative reports, and billing records.⁹ The eating disorder diagnosis could have been made any time, including prior to 2010, however all patients had an outpatient GI clinic visit within the specified 10-year period. Patient charts were reviewed to confirm eating disorder diagnosis, and patients were excluded if the eating disorder diagnosis could not be confirmed, if there was no outpatient clinical GI encounter (i.e., GI encounter occurred only for a procedure or while patient was hospitalized), or age at GI encounter was <18 years old.

Eating disorder diagnoses were defined using the Diagnostic and Statistical Manual of Mental Disorders eating disorder criteria, fourth and fifth editions (DSM-IV and DSM-5).¹⁰ As BED, OSFED and ARFID were added as diagnoses in the DSM-5 in 2013, these eating disorder diagnoses were grouped with EDNOS for our analyses.

GI Diagnosis Coding

Two coders (MNA, IGF) were trained to systematically review medical records for demographics (sex, age, and body mass index (BMI)) at the time of the GI encounter as well as GI diagnoses and eating disorder diagnoses as below: Coders reviewed the outpatient GI consult note and subsequent progress notes to determine the diagnosis conferred by the provider. For n=23, no definitive GI diagnosis was able to be confirmed (e.g., diagnosis listed as “abdominal pain”). For these patients, coders used Rome IV criteria to determine if a disorder of gut-brain interaction (DGBI) was present. Unclear diagnoses were reviewed by a gastroenterologist (KS) masked to eating disorder status.

Coders further categorized GI diagnoses as either functional/motility disorders, structural/inflammatory disorders, or unclassified, as described in Supplemental Methods.

Three patients had diagnoses related to family history of GI cancer without any GI symptoms or diagnoses. This group was categorized as unclassified diagnoses and excluded from the analysis.

Timeline of Eating Disorder Relative to GI Consult

To assess whether the eating disorder diagnosis was made before the GI consult encounter (preexisting eating disorder) or after the GI consult encounter (subsequent eating disorder), coders reviewed notes to determine the timing of the initial eating disorder diagnosis in relation to the date of GI encounter. If there was no definitive date documented in a note, but an approximate time or patient age when eating disorder symptoms developed was described in a note (e.g., “during college years”), this general time frame was compared to the GI encounter date. Patients with no documentation of timing of eating disorder diagnosis (n=7) were removed from further analysis.

Eating Disorder Diagnosis Coding for Remission Status

To assess presence of eating disorder symptoms at the time of GI consult, coders were trained by a clinical psychologist (HBM) to systematically review medical records for evidence of DSM-5 eating disorder criteria. Coders independently reviewed each patient’s medical record, focusing on notes from primary care, GI, psychology/psychiatry, nutrition, and social work encounters in the three months before and after the GI consult encounter. Coders completed a DSM-5 diagnostic checklist for AN, BN, BED, ARFID, and OSFED for each patient (Supplemental Table 1). Coders conferred “definite eating disorder” if patients fulfilled the entire criteria for any eating disorder, and “potential eating disorder” if patients fulfilled some, but not all, of the DSM-5 diagnostic criteria for an eating disorder. An independent blinded coder (IEW) reviewed the patients’ chart and all discrepancies in eating disorder diagnoses between original and blinded coding were then reviewed with a clinical psychologist (HBM) for final coding determination.

To determine remission status for patients with preexisting eating disorders at GI consult, the categorization from the DSM-5 diagnostic checklist was used.¹⁰ Patients who were categorized as “definite eating disorder” or “potential eating disorder” were considered not in remission at their GI encounter, while patients who did not meet any DSM-5 diagnostic criteria were considered in remission. Patients with unknown remission status (n=1) were excluded from analyses. For patients with subsequent eating disorders, the DSM-5 diagnostic checklist was used to evaluate for the presence of eating disorder symptoms at the time of GI consult.

Statistical Analysis

Five separate cohorts were analyzed: all patients, patients with preexisting eating disorders, patients with preexisting eating disorders in remission, patients with preexisting eating disorders not in remission, and patients with subsequent eating disorders. Within each group, continuous variables were summarized using means and standard deviations, while categorical variables were expressed as proportions. For discrete outcomes, univariate comparisons were made using a chi-square test and continuous outcomes were compared using an independent-sample t test.

In the cohort with preexisting eating disorders, we performed a multivariate logistic regression to evaluate the relationship between presence of a functional/motility GI diagnosis and a priori predictors: eating disorder diagnosis (AN, BN, EDNOS), age at time of GI consult, sex, BMI, and number of GI diagnoses conferred at consult. We also used multivariate logistic regression to evaluate the relationship between eating disorder remission status (coded at the time of GI consult) and a priori predictors including GI diagnosis category (structural vs functional/motility), age at time of GI consult, sex, BMI, and number of GI diagnoses conferred at consult.

In the subsequent eating disorder cohort, we performed a descriptive analysis of the patients with potential undiagnosed eating disorders at the time of GI consult.

Statistical analyses were performed using SAS Institute. Statistical significance was defined by a p-value of 0.05. This study was approved by the Institutional Review board at Massachusetts General Hospital (Protocol #2018P000504).

Results

We identified a total of 610 patients from our database, of which 344 met inclusion criteria with a GI consult and an eating disorder diagnosis (Figure 2). The population was predominantly female (91.9%) with a mean age of 37.2±11.6 years and a mean BMI of 26.4±8.9 kg/m². 134 patients (38.9%) were classified as having a diagnosis of EDNOS, 144 patients (33.1%) had AN, and 96 patients (27.9%) had BN (Table 1). Patients with AN and BN tended to be younger than those with EDNOS with mean BMI decreasing from EDNOS to BN to AN.

Figure 2.

Cohort breakdown

Note: Remission status was determined through application of DSM-5 criteria checklist (Supplemental Table 1). EDNOS: eating disorder not otherwise specified; AN: anorexia nervosa; BN: bulimia nervosa

Table 1.

Demographics of overall cohort of patients with an eating disorder diagnosis and GI consult (N=344)

	Overall n=344	Anorexia Nervosa n=114	Bulimia Nervosa n=96	EDNOS n=134	p-value
Patient Characteristics
Female sex, n (%)	316 (91.9)	108 (94.7)	93 (96.9)	115 (85.8)	0.004
Age at GI Consult, mean ± SD	37.2 ± 11.6	34.7 ± 10.9	35.6 ± 10.9	40.5 ± 11.9	0.0001
Body Mass Index at GI Consult, mean ± SD	26.8 ± 8.9	21.6 ± 5.2	26.0 ± 7.6	31.5 ± 9.6	<0.0001
GI Diagnosis Category
Structural GI Diagnoses, n (%)	144 (41.9)	45 (39.5)	33 (34.4)	66 (48.3)	0.06a
Functional/Motility GI Diagnoses, n (%)	197 (57.3)	67 (58.8)	62 (64.6)	68 (50.7)
Unclassified, n (%)	3 (0.9)	2 (1.8)	1 (1.0)	0 (0.0)
Relationship of Eating Disorder Diagnosis to GI Consult
Preexisting Eating Disorder Diagnosis, n (%)	255 (74.2)	100 (87.7)	82 (85.4)	73 (54.5)	<0.0001 a
Subsequent Eating Disorder Diagnosis, n (%)	82 (23.8)	12 (10.5)	11 (11.5)	59 (44.0)
Unknown, n (%)	7 (2.0)	2 (1.8)	3 (3.1)	2 (1.5)

Note.

^aComparisons excluded unclassified/unknown category

Structural GI Diagnoses: Inflammatory Bowel Disease, celiac disease, non-alcoholic fatty liver disease, other structural (esophagitis, congenital cecal bascule, lap band maladjustment, parasites, helicobacter pylori, hemorrhage of the rectum and anus, polypectomy bleed, infectious gastroenteritis, small bowel obstruction, rectal prolapse, hyperthyroidism, other ascites due to malnutrition), other-liver (alcohol induced steatosis, hepatoxicity, drug induced liver injury, viral hepatitis, hepatic sarcoidosis, hepatic dilation and liver fibrosis)

Functional/Motility GI Diagnoses: Irritable bowel syndrome, functional dyspepsia/gastroparesis, gastroesophageal reflux disease, constipation, other-functional/motility (dysphagia, globus sensation, esophageal spasm, achalasia, chronic nauseas, rumination syndrome, chronic abdominal pain, small intestinal bacterial overgrowth, sphincter of Oddi dysfunction, functional diarrhea)

GI Diagnoses among the Entire Cohort

Functional/motility GI disorders made up the majority of diagnoses overall (57.3% vs. 41.9% structural, and 0.9% unclassified) with IBS, functional dyspepsia/gastroparesis, and constipation representing the first, second, and fourth most common diagnoses, respectively (Figure 3). IBS was the single most common GI diagnosis in the cohort, diagnosed in 20.0% of patients overall and in 23.7% of patients with AN. GERD was the most common diagnosis among patients with BN (19.8% of patients), and “other structural diseases” (e.g. eosinophilic esophagitis, H. pylori) was the most common diagnostic category among patients with EDNOS. There were no differences in the proportion of functional/motility vs. structural disorders when comparing the different eating disorder diagnoses (58.8% functional/motility in AN, 64.6% in BN, and 50.8% in EDNOS, p=0.10).

Figure 3.

Heatmap with breakdown of GI diagnoses among (A) overall cohort of patients with a diagnosed eating disorder and GI consult, (B) patients with preexisting diagnosed eating disorder prior to GI consult and (C)patients with subsequent eating disorder diagnosed after GI consult (C)

IBS: Irritable Bowel Syndrome; FD: Functional Dyspepsia; GERD: Gastroesophageal Reflux; IBD: Inflammatory Bowel Disease; NAFLD: Non-alcoholic Fatty Liver Disease

Structural GI Diagnoses: Inflammatory Bowel Disease, celiac disease, non-alcoholic fatty liver disease, other structural (esophagitis, congenital cecal bascule, lap band maladjustment, parasites, helicobacter pylori, hemorrhage of the rectum and anus, polypectomy bleed, infectious gastroenteritis, small bowel obstruction, rectal prolapse, hyperthyroidism, other ascites due to malnutrition), other-liver (alcohol induced steatosis, hepatoxicity, drug induced liver injury, viral hepatitis, hepatic sarcoidosis, hepatic dilation and liver fibrosis)

Functional/Motility GI Diagnoses: Irritable bowel syndrome, functional dyspepsia/gastroparesis, gastroesophageal reflux disease, constipation, other-functional/motility (dysphagia, globus sensation, esophageal spasm, achalasia, chronic nauseas, rumination syndrome, chronic abdominal pain, eating disorder, small intestinal bacterial overgrowth, sphincter of oddi dysfunction, functional diarrhea

Eating Disorder Diagnosis Relative to GI Consult

We further examined the timing of eating disorder diagnoses relative to GI consult. 255 (74.2%) patients presented to GI after having already received an eating disorder diagnosis (preexisting eating disorder) whereas 82 (23.8%) were diagnosed after their GI consult (subsequent eating disorder).

Preexisting versus Subsequent Eating Disorder Cohorts

In the preexisting eating disorder diagnosis group, there was a greater proportion of patients with AN compared to the subsequent eating disorder diagnosis group (Table 2). Compared to the subsequent eating disorder group, the preexisting eating disorder group were more likely to be younger, female, and have a lower mean BMI (Table 2).

Table 2.

Demographics and GI diagnosis among patients with preexisting eating disorder at GI consult and subsequent eating disorder (n=337)

		Preexisting Eating Disorder n=255	Subsequent Eating Disorder n=82	p-value
Demographics
Females (%)		243 (94.9)	68 (82.9)	0.0005
Average Age at GI Consult (SD)		36.4 (11.2)	40.5 (12.1)	0.004
Average Body Mass Index at GI Consult (SD)		25.8 (8.7)	30.3 (8.8)	0.0001
Eating Disorder Subtype
Anorexia Nervosa, n (%)		100 (39.1)	12 (14.6)	<0.001
Bulimia Nervosa, n (%)		82 (32.0)	11 (13.4)
EDNOS, n (%)		73 (28.5)	59 (72.0)
GI Diagnoses
Structural GI Diagnoses, n (%)		94 (36.9)	46 (56.1)	0.001
Functional/motility GI Diagnoses, n(%)		159 (62.4)	35 (42.7)
Unclassified, n(%)		2 (0.8)	1 (1.2)
GI Diagnoses by Eating Disorder
Anorexia Nervosa	Structural GI Diagnosis, n (%)	37 (37.0)	7 (58.3)	0.09
Functional/motility GI Diagnosis, n (%)		62 (62.0)	4 (33.3)
Unclassified, n (%)		1 (10.00)	1 (8.3)
Bulimia Nervosa	Structural GI Diagnosis, n (%)	25 (30.5)	5 (45.5)	0.33
Functional/motility GI Diagnosis, n (%)		56 (68.3)	6 (54.6)
Unclassified, n (%)		1 (1.2)	0 (0.0)
EDNOS	Structural GI Diagnosis, n (%)	32 (43.8)	34 (57.6)	0.16
Functional/motility GI Diagnosis, n (%)		41 (56.2)	25 (42.4)
Unclassified, n (%)		0 (0.0)	0 (0.0)

Note. n=7 were excluded from analysis due to unknown timeline of eating disorder diagnosis in relation to GI consult. EDNOS: Eating disorder not otherwise specified.

GI Diagnoses in Preexisting versus Subsequent Eating Disorder Cohorts

Overall, functional/motility GI diagnoses were significantly more common the preexisting eating disorder group compared to the subsequent eating disorder group (62.5% vs 42.7%, p<0.01). In the preexisting eating disorder group, functional/motility GI disorders accounted for the top 3 diagnoses: IBS (23.4%), FD/gastroparesis (18.0%), and constipation (16.8%) (Figure 3). In the subsequent eating disorder group, the most common GI diagnoses at consult visit were “other structural” diagnoses (22.0%), followed by GERD (14.6%), and inflammatory bowel disease (12.2%).

In a multivariable logistic regression evaluating odds of a functional/motility GI disorder among those with preexisting eating disorders, younger age at GI consult, female sex and number of different GI diagnoses at time of GI consult were associated with a functional/motility GI diagnosis (Table 4).

Table 4.

Multivariate analysis predicting functional/motility GI diagnosis in the overall cohort and predicting eating disorder remission at time of GI consult

Predicting Functional/Motility GI Diagnosis
	OR (95%CI)	p-value
Age at consult by decade	0.71 (0.56 – 0.91)	0.006
Female Sex	2.96 (1.09 – 8.00)	0.03
Eating disorder subtype
EDNOS	Reference	-
Anorexia Nervosa	0.82 (0.40 – 1.68)	0.41
Bulimia Nervosa	1.13 (0.57 – 2.23)	0.47
Body Mass Index
Normal	Reference	-
Underweight	1.87 (0.73 – 5.79)	0.05
Overweight	0.70 (0.32 – 1.52)	0.28
Obese	0.61 (0.30 – 1.23)	0.09
GI Diagnoses
Number of GI Diagnoses at GI Consult	2.14 (1.23 – 3.70)	0.007
Predicting Remission Status at GI Consult
Age at consult by decade	1.27 (0.93 – 1.73)	0.12
Female Sex	0.40 (0.08 – 2.08)	0.28
Eating disorder subtype
EDNOS	Reference	-
Anorexia Nervosa	1.52 (0.62 – 3.72)	0.77
Bulimia Nervosa	2.85 (1.19 – 6.80)	0.02
Body Mass Index
Normal	Reference	-
Underweight	0.13 (0.04 – 0.46)	<0.0001
Overweight	1.93 (0.75 – 4.96)	0.007
Obese	1.07 (0.44 – 2.58)	0.25
GI Diagnoses
Number of GI Diagnoses at GI Consult	0.047 (0.26 – 0.85)	0.01
Functional/motility diagnosis	0.64 (0.22 – 1.24)	0.20

Note. EDNOS: eating disorder not otherwise specified

Remission Status in Preexisting Eating Disorder Cohort

Of the 255 patients with a preexisting eating disorder, 113 (44.3%) were not in remission from their eating disorder (by DSM-5 coding) at the time of GI consult. The highest rate of non-remission was in those with preexisting AN and lowest in preexisting BN (Table 3). Compared to those in remission at GI consult, those who were not in remission were younger and had a lower BMI (Table 3). 26 (23.0%) patients with a preexisting eating disorder were considered not in remission at GI consult due to meeting DSM-5 criteria for ARFID at GI consult; of these 26 patients, 16 (61.5%) patients previously were diagnosed with AN, 6 (23.1%) patients were previously diagnosed with EDNOS, and 4 (15.4%) patients were previously diagnosed with BN.

Table 3.

Demographics and GI diagnoses by functional/motility and structural categories among patients in remission and not in remission (n= 254)

		In Remission n=141	Not in Remission n=113	p-value
Demographics
Females (%)		133 (93.7)	109 (96.5)	0.31
Age at GI Consult, mean ± SD		37.8 ± 10.8	34.5 ± 11.4	0.02
Body Mass Index at GI Consult, mean ± SD		26.9 ± 8.4	24.5 ± 8.8	0.04
Eating Disorder Subtype
Anorexia Nervosa, n (%)		48 (33.8)	52 (46.0)	0.01
Bulimia Nervosa, n (%)		56 (39.4)	25 (22.1)
EDNOS, n (%)		37 (26.1)	36 (31.9)
GI Diagnoses
Structural GI Diagnoses, n (%)		62 (43.7)	31 (27.4)	0.005 a
Functional/motility GI Diagnoses, n (%)		77 (54.6)	82 (72.6)
Unclassified, n (%)		2 (1.4)	0 (0.0)
GI Diagnoses by Preexisting Eating Disorder Diagnosis
Anorexia Nervosa	Structural GI Diagnosis, n (%)	22 (45.8)	15 (28.9)	0.07a
Functional/motility GI Diagnosis, n (%)		25 (52.1)	37 (71.7)
Unclassified, n (%)		1 (2.1)	0 (0.0)
Bulimia Nervosa	Structural GI Diagnosis, n (%)	19 (33.9)	5 (20.0)	0.20a
Functional/motility GI Diagnosis, n (%)		36 (64.3)	20 (80.0)
Unclassified, n (%)		1 (1.8)	0 (0.0)
EDNOS	Structural GI Diagnosis, n (%)	21 (33.9)	11 (30.6)	0.02 a
Functional/motility GI Diagnosis, n (%)		16 (64.3)	25 (69.4)
Unclassified, n (%)		0 (0.0)	0 (0.0)

Note. N=1 patient with pre-existing ED had unknown remission status due to lack of information and was not included in analysis. EDNOS: Eating disorder not otherwise specified.

^aUnclassified GI Diagnosis categories were not included in statistical analyses

Patients not in remission were more likely to be diagnosed with a functional/motility GI disorder compared to those in remission (72.6% with functional/motility diagnoses in non-remission vs 54.9% in remission, p=0.01). IBS was the most prevalent diagnosis among both those in and not in remission (Figure 4). 84.6% of the 26 patients meeting DSM-5 ARFID criteria at GI consult were diagnosed with a functional/motility GI disorder.

Figure 4.

Breakdown of GI diagnoses by remission status

IBS: Irritable Bowel Syndrome; FD: Functional Dyspepsia; GERD: Gastroesophageal Reflux; IBD: Inflammatory Bowel Disease; NAFLD: Non-alcoholic Fatty Liver Disease

Remission status was determined by evaluating patient charts and applying the DSM-5 criteria

Color ranges from green to red, with green being lower proportion of GI diagnoses and red being greater proportions

In a multivariable logistic regression evaluating the odds of being in remission at GI consult, preexisting BN was associated with increased odds of remission, while being underweight, and increasing number of diagnoses at GI consult were associated with decreased odds of remission (Table 4).

Undiagnosed Eating Disorders at the Time of GI Consult (Subsequent Eating Disorders)

Among the 82 patients who had subsequent eating disorder diagnoses, we found 14 (17.1%) patients met DSM-5 criteria for an active, undiagnosed eating disorder at GI consult (by DSM-5 coding): 2 with AN, 7 with ARFID, and 5 with OSFED. These patients were later diagnosed with EDNOS (10/14), AN (3/14) and BN (1/14) by DSM-IV. Among the 14 patients, the average time to an eating disorder diagnosis after the GI consult was 3.8 years (range=1.0–8.3 years).

Discussion

In this retrospective study of patients with eating disorders and outpatient GI encounters over a 10-year period, we found that most patients were diagnosed with an eating disorder prior to their GI consult, that these patients were more likely to have functional/motility GI disorders, and almost half were not in remission from their eating disorder at the time of GI consult. These findings help to clarify the complex, bidirectional relationship between GI symptoms, eating disorders, and remission status, and may provide insight into the GI provider’s role in caring for patients with eating disorders.

Eating disorders preceded GI consults for over 75% of patients in our study, which is consistent with previous studies assessing GI symptom onset and persistence in eating disorders.^11,12 One retrospective study in IBS found that eating disorder symptoms preceded IBS diagnosis for the majority (87%) of patients, with a mean duration between eating disorder and IBS onset of 10.4 years.¹³ Another longitudinal study found that 34% of patients who were hospitalized for eating disorder developed a new DGBI one year later.¹⁴ While the timing of GI encounters may not exactly reflect onset of GI symptoms, it is conceivable that eating disorders lead to GI symptoms through their impact on the enteric, central, and autonomic nervous systems.^12,15–17 Eating disorders and their sequalae (e.g. weight loss, malnutrition) have been shown to result in pathologic changes in GI motor function (e.g. delayed stomach emptying or delayed colonic transit)¹⁸ and associated GI symptoms (Figure 1, panel A). While some studies show that these altered processes improve with weight restoration and cessation of compensatory behaviors (e.g. purging),¹² GI symptoms may develop or persist even when eating disorders are in remission (Figure 1, panel B).^11,19 The exact reason for this delayed onset is unknown, but may be due to long-term neurosensory consequences, possibly mediated at the level of the microbiome.²⁰

Among patients with a preexisting eating disorder diagnosis at GI consultation, we found that functional/motility diagnoses were more common than structural diagnoses, with over half of this cohort diagnosed with IBS, FD/gastroparesis, and constipation alone. The relationship between DGBIs and eating disorders is well-established in the literature, with studies showing high rates of co-occurrence^7,21,22 and increased eating disorder pathology associated with both number of DGBI diagnoses²³ and DGBI symptom severity.^6,24 IBS, in particular, is the most common GI diagnosis associated with disordered eating,²³ reported in up to 93% of outpatients with AN with increasing prevalence associated with both lower BMI and longer duration of eating disorder symptoms.²⁴ Our understanding of the pathophysiologic underpinnings of DGBI is still evolving– but proposed mechanisms include visceral hypersensitivity, dysmotility, intestinal dysbiosis, and altered mucosal permeability,²⁵ all of which may be affected by diet restrictions and compensatory behaviors (e.g. purging) present in eating disorders.²⁶

Conversely, patients with a subsequent eating disorder diagnoses were less likely to have a functional/motility disorder, with the most common diagnoses of “other-structural” (e.g., esophagitis, infection) followed by GERD and inflammatory bowel disease. Onset of eating disorder symptoms following these diagnoses raises the question of whether GI disorders and their management can lead to disordered eating (Figure 1, panel D).²⁷ Patients with GI symptoms commonly associate their symptoms with food and may develop dysfunctional illness beliefs regarding their disease and eating-related consequences.^15,28 One review assessing disordered eating in chronic GI conditions (including celiac disease, IBS, and IBD) found that the majority of the 9 included studies showed that disordered eating behavior is more prevalent in patients with chronic GI conditions than the general population and may be associated with GI symptoms and anxiety.²⁸ In IBD specifically, one study found that 13% of patients screen positively for disordered eating, which was associated with having IBD diagnosed during childhood.²⁷ GI symptom management may also contribute to eating disorder development, and patients with diet-controlled GI disease or who use exclusion diets as treatment for GI symptoms may be at particular risk.²⁹ Although this has not been explored prospectively, one retrospective study found that those with a history of trying an exclusion diet were over three times as likely to have ARFID symptoms.³⁰

Importantly, we found that almost half (44.3%) of patients with a preexisting ED diagnosis still had an eating disorder (coded by DSM-5) at their GI consult visit. Additionally, among the 82 patients who did not have a chart ED diagnosis until after the GI consult, 14 met eating disorder criteria (coded by DSM-5) at the visit (Figure 1, panel C). This finding underscores the importance of eating disorder screening among GI clinicians, which is rarely part of a standard GI intake.¹⁵ Asking about eating disorder symptoms in a non-judgmental way can foster open communication about restricted eating practices and other behaviors,³¹ and patients with active eating disorder symptoms may benefit from understanding how their GI symptoms impact their eating behaviors and may warrant different GI treatment strategies.^32,33 In addition to avoidance of exclusion diets, other treatment strategies, such as cathartic laxative agents, may not be appropriate for patients with active eating disorders.^31,34 This seems to be especially true for patients who are underweight at GI consult, as our data suggest that these patients were 87% less likely to be in remission. Patients with several GI diagnoses were also less likely to be in remission from their eating disorder at GI consult visit, suggesting that worse GI symptomatology may impact eating disorder remission status. Addressing GI symptoms in conjunction with eating disorder specific therapy may help to increase treatment success, as ongoing symptoms may reinforce maladaptive behaviors.³¹

This study has several limitations due to its retrospective design. Notably, our cohort was assembled by selecting patients with chart diagnoses of eating disorders, which likely does not adequately capture all eating disorder patients. Our cohort was predominantly female sex (91.9%) which limits generalizability and also does not adequately reflect the demographics of eating disorders, with some studies reporting that males make up 25–33% of eating disorder patients.³⁵ By retrospectively evaluating GI encounter-associated diagnoses rather than patient-reported symptoms, it is possible that we did not gain a full understanding of the temporal relationship between GI symptoms and eating disorders—some GI symptoms may have preceded the GI encounter, or even the eating disorder itself. Our categorization of eating disorder remission status was also limited by reliance on free-text documentation in notes, and out of the 113 patients not in remission, we were only able to definitively confer an eating disorder diagnosis for 55 patients.

During our study time frame (2010 to 2020), the DSM-5 was released with a notable addition of ARFID. Unlike other eating disorders, food avoidance and restriction in ARFID is not motivated by shape/weight concerns, but instead driven by low appetite/lack of interest in eating, sensory sensitivities to food or fear of aversive consequences of eating.¹⁰ ARFID has garnered attention in the GI field due to the symptom overlap with many GI conditions, with one study showing over 50% of patients with IBD, achalasia, celiac disease and eosinophilic esophagitis met ARFID screening criteria.³⁶ ARFID rates in GI disorders may be overinflated due to appropriate, adaptive dietary restrictions—but in a subset of patients, these restrictions become pathologic and result in medical or psychosocial consequences.³³ Because ARFID was first described in 2013 (within our study timeframe of 2010–2020), these patients were included within the EDNOS cohort, and therefore this study lacks specific data on this important subset of patients. We did find that among the subsequent eating disorders diagnoses, most (59/82) were classified as EDNOS (with ARFID explicitly diagnosed in 5 of the 59 patients)– which likely reflects growing clinical awareness.

More prospective research on comorbid eating disorder and GI symptom development is certainly needed to truly understand the relationship between eating disorders and GI disorders – but our findings suggest that eating disorders may precede GI assessment, particularly in patients with functional/motility disorders. GI providers should incorporate eating disorder assessment as an essential component of the intake, as the thoughtful management of GI symptoms in the context of eating disorders is crucial to achieving and maintaining remission.