Laxative abuse is associated with a depleted gut microbial community structure among females and males with binge-eating disorder or bulimia nervosa: The Binge Eating Genetics Initiative (BEGIN)

Abstract

Objective

To assess the associations of binge eating, compensatory behaviors, and dietary restraint with the composition and diversity of the intestinal microbiota among participants with binge-eating disorder (BED) or bulimia nervosa (BN).

Methods

We analyzed data from 265 participants aged 18–45 years with current BED or BN enrolled in the Binge Eating Genetics Initiative (BEGIN) study. We evaluated the associations of binge-eating frequency, presence/absence and frequency of vomiting, laxative use, and compulsive exercise, and dietary restraint with abundances of gut microbial genera, species, and diversity (Shannon diversity, Faith phylogenetic diversity, and Peilou’s evenness). General linear models adjusted for potential confounders, including age and current BMI, modeled associations; p-values were corrected for the false discovery rate.

Results

The normalized abundance of four genus- and species-level gut microbes and three diversity indices were lower among BEGIN participants who reported any laxative use compared to those who reported no laxative use. Vomiting frequency was positively associated with the normalized abundance of genus Escherichia-Shigella, a potential pathobiont, although the association was attenuated to non-significance after adjustment for age, BMI, and binge-eating episodes.

Conclusions

Laxative use was highly and uniformly predictive of a reduced gut microbial diversity including of potential commensals and pathobionts and should be assessed and accounted for in all future microbial studies of eating disorders. Future studies should collect data on specific medications—especially laxatives—and dietary intake to obtain unbiased estimates of the effect of eating disorders on the gut microbiota and identify potential downstream clinical implications.

Registration:

ClinicalTrials.gov identifier: NCT04162574

Introduction

Binge-eating disorder (BED) and bulimia nervosa (BN) are serious psychiatric conditions characterized by the hallmark feature of binge-eating behavior (consuming an objectively large amount of food with a concomitant sense of loss of control) and, in the case of BN, by compensatory behaviors (primarily vomiting, laxative use, and compulsive exercise) with the objective of counteracting the effects of a binge and/or inducing weight loss. Both disorders are associated with high psychiatric and somatic comorbidity (1–3) and elevated suicide risk (4–7). The etiology of these conditions is complex and poorly understood.

First line treatments for BED and BN include cognitive behavioral therapy (CBT) (8), interpersonal psychotherapy (IPT) (9), and medication management. Multiple studies have shown that fluoxetine, lisdexamfetamine dimesylate, or administration of these medications in combination with CBT can reduce binge-eating and vomiting episodes compared to placebo among adults with BED or BN, although symptoms do not fully resolve and treatment effects may not be long-term (10–19). Thus, novel strategies that can support treatment effectiveness and prevent relapse over the long-term are critically needed for individuals suffering from these eating disorders.

The concept of a microbiota-gut-brain axis refers to the cross-talk that occurs amongst the intestinal microbiota (i.e., the community of gut microbes residing in the intestinal tract), the gut, the enteric innate immune system, neurons, and the brain (20). The collective gut microbial genomes in the human colon are vast (21) and reflect the microbial enzymatic capabilities to oxidize host substrates such as ingested nutrients or colonic mucins (22) for their own metabolic needs. The relative proportions of gut microbial taxa impact the quantity and types of microbial byproducts that are produced through utilization of host substrates, including essential nutrients (23) and microbial metabolites (24–28). These metabolites—predominantly short-chain fatty acids and secondary bile acids, among others—are sensed by the human host through vagal mechanisms or through their direct secretion into host circulation (20). The downstream host response to these microbial metabolites may take the form of changes in satiety hormone production and energy metabolism (29), mood, sympathetic nervous system activation, and immune modulation (20), all of which may have implications for eating disorder prognosis. Thus, the microbiota-gut-brain axis is critical to examine in people with BED and BN.

Gut microbial composition has been investigated in a limited number of studies in patients with anorexia nervosa (AN), with multiple reports of a reduced intestinal microbial diversity and abundance of beneficial intestinal microbes, (i.e., those that produce short-chain fatty acids such as butyrate), compared to controls without an eating disorder (30–32). However, to our knowledge, only one such study has been conducted in individuals with BED, which found that the relative abundances of 11 gut microbial genera differentiated study participants with and without BED (33). No such studies have been conducted in individuals with BN, to our knowledge.

The composition of the intestinal microbiota changes rapidly following major shifts in diet (34) and is therefore likely influenced by multiple eating behavioirs characteristic of BED and BN, including binge eating, compensatory behaviors (in particular, laxative abuse (35), i.e., taking laxatives for the purpose of controlling one’s weight or shape), and dietary restraint—a restrictive behavior characteristic of AN that also occurs in BED and BN (36). Therefore, the objective of this study was to evaluate the associations of 1) binge-eating frequency, 2) the presence and frequency of various compensatory behaviors (i.e., vomiting, laxative use, and compulsive exercise), and 3) self-reported measures of dietary restraint with the genus- and species-level taxonomy and intestinal microbial diversity of actively ill participants with self-reported BED or BN.

Methods

Study sample

Procedures for the Binge Eating Genetics Initiative (BEGIN) study have been described elsewhere (37). In brief, the BEGIN study recruited 1,166 men and women ages 18–45 with BED or BN between August 2017 and August 2021. BEGIN-US participants had to meet Diagnostic and Statistical Manual of Mental Disorders, 5^th Edition (DSM-5) (38) criteria for BED or BN (via the ED100K.v2 questionnaire, described below) and were required to have an existing iPhone 5 or later where they interacted with the Recovery Record app for 30 days (39). Individuals were excluded if they were currently pregnant or breastfeeding, had a history of bariatric surgery (due to influence on appetite), were currently using hormones, had been hospitalized for eating disorder treatment in the two weeks prior to enrollment, displayed suicidality at screening, or had used antibiotics or probiotics in the past 30 days (due to collection of fecal samples for assessment of the intestinal microbiota). The research was reviewed and approved by the University of North Carolina at Chapel Hill Biomedical Institutional Review Board. Study participation began after participants signed informed consent.

All participants were asked to complete the ED100K.v2 (40), which is a validated self-report questionnaire assessing lifetime history of BED, BN, and AN based on DSM-5 criteria, prior to study enrollment. Information about age, sex, race, ethnicity, and body mass index (BMI) was also collected. Once enrolled, all were asked to complete baseline assessments which included the Eating Disorder Examination Questionnaire, version 6 (EDE-Qv6) (41). In addition, the number of binge-eating episodes was taken from the item which references “eating an unusually large amount of food (given the circumstances)”: “Over the past 28 days,… On how many of these times did you have a sense of having lost control over your eating (at the time you were eating)?” Participants who enrolled earlier in the study were also asked to provide a single fecal sample using a uBiome (formerly uBiome, Inc.) home kit¹, which was mailed directly to them. Participants then sent their samples to uBiome for processing. These kits are stable at room temperature, so no additional shipping requirements were necessary.

For the current study, we included participants who indicated that they were currently binge eating (who responded, “I still currently have regularly occurring overeating episodes” to the question “How old were you when the regular episodes of binge eating stopped?”), had at least four binge-eating episodes in the month prior to study enrollment, and submitted a gut microbiome sample which was processed and passed quality control thresholds.

From the 1,166 participants enrolled in BEGIN, we received uBiome gut microbiota sequence data for 396 samples, prior to uBiome Inc. closure. We were unable to match the uBiome ID number with a BEGIN ID number for three samples and two samples with duplicate IDs were removed, leaving 391 unique samples. Thirty-five participants were further excluded due to missing the EDE-Qv6 in its entirety, and another 35 were excluded for not meeting all criteria for lifetime history of BN or BED, leaving 321 participants. Following quality control steps for analysis of 16S rRNA gene sequence reads, an additional seven samples were excluded due to an insufficient number of gene sequence reads (<1,000), leaving 314 samples. Finally, 41 participants who reported <4 episodes of binge eating in the past 28 days according to the EDE-Qv6 and eight participants with a BMI classified as being underweight (<18.5 kg/m2) including seven with a lifetime history of AN were excluded. We made this decision on the basis that gut microbial composition and intestinal microbial diversity vary by weight status and between individuals with and without AN (42). Median BMI was 30.4 (IQR 24.9, 38.5) among included participants and 17.3 (IQR 17.0, 18.1) among those excluded for a low BMI. Furthermore, individuals with a low BMI at the time of study participation may have been engaging in more restrictive eating disorder behaviors than those without a low BMI, which could impact the composition and diversity of the intestinal microbiota (42, 43). These decisions align with the overarching objectives of the parent BEGIN study to extend genetic and microbiome research previously conducted in AN to individuals with BED and BN (37). Our final sample size consisted of 265 participants, each with one set of measurements.

Of note, the sample size for the parent BEGIN study was determined based on the primary aim of BEGIN, which was to collect an adequate number of samples to conduct a genome-wide association study of BN and BED (37). Given the exploratory nature of the present analysis, no formal power analysis was conducted.

Measures

EDE-Qv6

Self-reported disordered eating behaviors were assessed using the EDE-Qv6. This instrument was also used to evaluate current eating disorder diagnosis, in accordance with previously published algorithms (44). Briefly, participants who reported engaging in any compensatory behavior (vomiting, laxative use, or compulsive exercise) at least once weekly over the prior four weeks was classified as currently having BN; otherwise, the participant was classified as having BED. The EDE-Qv6 was used to derive our primary exposure variables of binge-eating frequency, the presence/absence and frequency of the three compensatory behaviors (vomiting, laxative use, and compulsive [i.e., driven] exercise), and dietary restraint. Dietary restraint was calculated by averaging EDE-Qv6 items 1–5. Ten participants reported 1 vomiting episode but were not classified as having BN because they reported <1 weekly episode of a compensatory behavior in the prior four weeks. There was one such participant for laxative use (reported only 1 episode of laxative use in the prior 4 weeks) and 30 such participants for compulsive exercise (all reported only 1 episode of compulsive exercise in the prior 4 weeks). In total, 39 participants who did not meet the classification for BN reported at least one episode of a compensatory behavior in the prior 4 weeks. Therefore, these participants were included in the linear regression analyses in which microbial diversity and taxonomic abundance was regressed from the frequency of each compensatory behavior.

We elected to use the dietary restraint subscale given that it asks about eating behaviors specifically and is therefore likely to be related to the composition and diversity of the intestinal microbiota. On the other hand, we believe that the eating concern, weight concern, and shape concern EDE-Qv6 subscales are less likely to reflect eating behaviors that can impact intestinal microbial composition or diversity and we therefore did not use these subscales in analysis.

Gut Microbiota Characterization

DNA Isolation, Amplification, and 16S rRNA Gene Sequencing

DNA isolation and amplification of the 16S rRNA gene were carried out via uBiome’s proprietary methods (45). 16S rRNA gene amplicons from each sample were individually barcoded and sequenced in multiplex on the NextSeq 500 platform in a 150bp paired-end modality.

Data Processing

Forward and reverse reads were retained if Q-scores > 30. Given that forward and reverse reads did not overlap and were too short to merge, only forward sequences were used in our analyses. In addition, leading random nucleotides were trimmed. Chimera sequences were removed using the VSEARCH uchime_denovo algorithm2. Forward and reverse reads that matched with at least 77% sequence identity to the same sequence in version 132 of the Divisive Amplicon Denoising Algorithm (DADA2)-formatted reference database SILVA were assumed to be 16S sequences (46). 16S rRNA gene sequences that passed these quality controls were managed in the Quantitative Insights Into Microbial Ecology (QIIME) 2 analysis pipeline. Sequences were demultiplexed and denoised via the DADA2 to generate sequence variants (47) at 100% identity threshold. The total number of sequence reads was 25,722,101 (85,851 [interquartile range, IQR, 54,520, 120,420] per sample) and total number of sequence variants (i.e., features, or unique gut microbial taxa) generated by DADA2 was 4,722.

Features with a frequency <0.01% summed across all samples were removed from analysis (46). We retained genus- and species-level features in analysis that were present in at least 25% of samples (i.e., non-rare taxa). Given the zero-inflated nature of gut microbial counts, which resulted in non-normal and heteroscedastic model residuals, we normalized the abundance of sequence reads belonging to the retained taxa according to a previously published equation (48):

$${\text{log}}_{10}\left(\left[\frac{\mathit{\text{Raw count in sample }}\left(i\right)}{\#\mathit{\text{of sequences in sample}}\left(i\right)}\times \mathit{\text{Average}}\#\mathit{\text{of sequences per sample}}\right]+1\right)$$

We used SILVA to perform taxonomic classification (46). For the analysis of intestinal microbial diversity, samples were rarefied to a sequencing depth of 4,000 to reduce the potential impact of sequencing bias (49). We selected four metrics of intestinal microbial diversity, which have complementary interpretations: the Chao-1 index, which estimates richness (i.e., the number of unique species) and can handle low abundance microbes (50); Shannon diversity, which increases with both increasing richness and evenness (i.e., how equally abundant the taxa are) (51); Faith phylogenetic diversity, which accounts for taxonomic interrelatedness (52); and Peilou’s evenness (53).

Demographics

Standard questionnaires collected self-reported age, race, ethnicity, and current BMI. Sex assigned at birth was asked in the feasibility phase and gender identity was asked in the main phase of the study. To reconcile these, single nucleotide polymorphism (SNP) tracing was performed on all participants who submitted a saliva sample across both phases. Genetic sex and reported sex assigned at birth matched in 98% of samples; SNP tracing results were inconclusive for the other 2%. Thus, sex for these participants was assigned as reported.

Statistical analysis

We constructed general linear models predicting the abundance (log-normalized counts) of each detected non-rare genus- and species-level microbe and each microbial diversity index from EDE-Qv6 variables: 1) binge-eating frequency; 2) the presence or absence of each compensatory behavior (laxative use, vomiting, or compulsive exercise); 3) the frequency of each compensatory behavior; and 4) dietary restraint score. We elected to focus on lower-order taxa as opposed to broader taxonomic classifications (i.e., phylum, family) to identify gut microbes linked with specific metabolic functions, which may serve as biomarkers of eating disorders and help to generate hypotheses for future research. This approach has been used in a prior observational cohort study of the gut microbiota (54). Linear specifications were used for each set of models given that inclusion of higher order terms was not found to improve model fit. We evaluated potential multicollinearity using the variance inflation factor, which was <5 in models testing associations of each exposure variable with the summary outcome of microbial diversity. Multicollinearity was therefore deemed to be negligible (55), which was further supported by weak bivariate correlations amongst the number of binge-eating and compensatory behavior episodes (Spearman’s rho=0.004–0.23, Table S1, Supplemental Digital Content). We winsorized one outlier value for vomiting frequency which had undue influence on the results to 10% greater than the next greatest value (56) and show the results before and after winsorization. Model 1 was unadjusted. Model 2 was adjusted for age and current BMI for all associations. In addition, Model 2 was adjusted for binge-eating frequency when a compensatory behavior (vomiting, laxative use, or compulsive exercise) was the exposure variable, and for the frequency of each compensatory behavior when binge eating was the exposure variable. We corrected for multiple hypothesis testing using the false discovery rate (FDR) procedure, which is designed to minimize type 1 error (i.e., false positives) and in doing so corrects p-values to produce q-values (57). Results were considered to be statistically significant at q<.05. All statistical analyses were conducted using SAS^® version 9.4. Data are available upon request.

Results

The 265 participants included in the present study had a median age of 29.0 years (IQR 24.0, 35.0) and a current BMI of 30.4 (IQR 24.9, 38.5 kg/m²) (Table 1). Approximately 87% reported being Caucasian and <5% reported identifying with each of the following: Asian, Black or African American, more than one race, Native American, other race. Six percent reported a Hispanic ethnicity. Genetic sex was reported as female for 217 (81.9%) participants and 48 (18.1%) were male. The majority of participants (189, 71.3%) reported a lifetime history of both BED and of BN; 21 (7.9%) participants reported a lifetime history of only BED; and 6 (2.2%) of participants reported a lifetime history of BN only or subthreshold binge-eating disorder only. According to the EDE-Qv6, all participants reported ≥4 binge-eating episodes (median 14 [IQR 8, 20]), approximately 43% reported ≥4 episodes of a compensatory behavior in the past 28 days, and 27.6%, 11.7%, and 48.3% of participants reported at least one episode of vomiting, laxative use, or compulsive exercise, respectively. The range of episodes of vomiting, laxative use, and compulsive exercise among participants with ≥1 episode was 1–100 (1–50 without the outlier of 100 episodes), 1–30, and 5–30, respectively. Median restraint score was 3.0 (IQR 2.0, 4.0).

Table 1.

BEGIN Participant Characteristics (n=265)

	Median (IQR) or N (%)
Age (years)	29.0 (24.0, 35.0)
Female genetic sex	217 (81.9)
Current BMI (kg/m2)	30.4 (24.9, 38.5)
Lifetime ED diagnosis
Binge-eating disorder (n=228)	225 (98.7)
Bulimia nervosa (n=248)	226 (91.1)
Anorexia nervosa (n=264)	37 (14.0)
Current Eating Disorder diagnosis (44)
Binge-eating disorder	118 (44.5)
Bulimia nervosa	73 (27.6)
Subthreshold bulimia nervosa	74 (27.9)
Eating Disorders Examination-Questionnaire1
Restraint	3.0 (2.0, 4.0)
Binge-eating episodes	14 (8, 20)
Vomiting	73 (27.6)
Vomiting episodes2,3	6 (3, 18)
Laxative use	31 (11.7)
Laxative use episodes2	5 (3, 10)
Compulsive exercise	128 (48.3)
Compulsive exercise episodes2	5 (3, 13)

¹The Eating Disorder Examination Questionnaire version 6 was used; items query the past 28 days.

²Descriptive statistics for compensatory behaviors were computed only among participants who reported at least one episode of a compensatory behavior, for each compensatory behavior type.

³Statistic includes outlier of 100 vomiting episodes. Without outlier, median (IQR) is 6 (3, 18).

We detected 174 genus- and species-level intestinal microbial taxa in the stool of our study participants with BED or BN. Of these, we retained 108 non-rare taxa.

According to both unadjusted and adjusted FDR-corrected estimates (Figure 1), participants who reported any laxative use had a reduced abundance of four different intestinal microbes or microbial groups (Eubacterium ventriosum, Alistipes, Bilophila, and GCA900066575) and reductions in three intestinal microbial diversity indices (Chao-1 Index, Faith phylogenetic diversity, and Shannon diversity) compared to participants who did not report laxative use (Figure 2, q<.05 for all). Per unadjusted estimates, vomiting frequency was positively associated with the normalized abundance of the genus Escherichia-Shigella (q=0.047). However, winsorization of the outlier value of 100 vomiting episodes attenuated the association to non-significance (q=0.07, Figure S1, Supplemental Digital Content), as did adjustment for age, current BMI, and binge-eating frequency with or without the outlier (crude q=0.17, adjusted q=0.27). We did not detect statistically significant associations of laxative use frequency, compulsive exercise, binge-eating frequency, or dietary restraint score with the abundance of any gut microbial taxa or diversity indices.

Figure 1. Heatmaps with standardized crude (A) and adjusted (B) beta estimates showing genus- and species-level intestinal microbes that were associated with the presence/absence or frequency (number of episodes) of each compensatory behavior.

Asterisks denote a statistically significant q-value <.05. Adjusted models included age, current BMI, and binge-eating frequency. (Color image is available online only at the Psychosomatic Medicine web site.)

Abbreviation: F—Frequency

Figure 2. Genus- and species-level gut microbes (A) and microbial diversity indices (B) that were whose relative abundance was lower among BEGIN participants who reported any laxative use compared to participants who reported no laxative use.

Q-values are from models adjusted for age, current BMI, and binge-eating frequency.

BEGIN participants included in the present analysis were similar to the larger study cohort. However, there was a slightly greater proportion of females and a greater number of binge-eating episodes among those excluded (Table S2, Supplemental Digital Content).

Discussion

In our sample of study participants who met DSM-5 criteria for BED or BN, we found reductions in the abundance of four gut microbial taxa and three intestinal microbial diversity indices among participants who reported any laxative use compared to those who reported no laxative use. Vomiting frequency was positively associated with the abundance of the genus Escherichia-Shigella, although the association was attenuated to non-significance following the winsorization of an extreme outlier and adjustment for potential confounding variables. As the Escherichia-Shigella group encompasses known gut pathogens, it is tempting to speculate that repeated self-induced vomiting may result in pathogenic infection through the depletion of one’s “normal” intestinal microbial defense systems (58); however, given the cross-sectional nature of this research, future studies with longitudinal designs should confirm the temporality of this association.

Laxative abuse is a pernicious symptom of eating disorders that has been associated with increased features of suicidality, self-harm, feelings of emptiness, anger, and borderline personality disorder (59), and with severe and potentially lethal electrolyte and acid-base changes (60). We found a global reduction in the abundance of several intestinal microbes and diversity metrics among participants who reported laxative use compared to those who did not. This included a reduction in the potentially beneficial short-chain fatty acid producers Alistipes and E. ventriosum. Additionally, we observed reductions in “pathobionts” (i.e., opportunistic pathogens (61)), Bilophila and GCA900066575 of the Firmicutes phylum, which have been associated with colitis and with cardiovascular risk factors in mice, respectively (62, 63).

Studies implicating the gut microbiota in human health are often conflicting, which makes interpretation difficult. For example, while the genus Alistipes encompasses gut microbes that can produce beneficial short-chain fatty acids which reduce host inflammation (64), Alistipes has also been associated with increased systolic blood pressure in humans (65). Such discrepancies may be explained by the fact that each gut microbial genus is comprised of numerous species and an even greater number of strains, which have unique and overlapping metabolic capabilities. Furthermore, the composition of an “ideal” or “healthy” gut microbiota has not yet been identified, is likely to be personalized, and must consider the abundances of each microbe relative to all others (66). Interestingly, the abundance of Alistipes was increased among 1,883 Dutch individuals who did not expressly have an eating disorder but reported laxative use, compared to those who did not use laxatives (67), although importantly, laxative type and frequency were not reported. It is possible that a high frequency of laxative use (as in the setting of BN) leads to reductions in this microbe, likely via changes in the intestinal environment such as pH and osmolality (67). Given that a small proportion (~11%) of our study participants reported laxative use, our results require replication in additional studies focused on individuals with BN.

Although nearly one-fifth our study participants were male, a larger sample of males—in whom compulsive exercise tends to feature more prominently—may be needed to detect associations between compulsive exercise and features of the intestinal microbiota (68). It may also be necessary to study individuals with BN separately, as compensatory behaviors are not a hallmark feature of BED.

The decreased microbial diversity among participants who reported laxative use is consistent with the effect of polyethylene glycol (PEG, an osmotic laxative) on the murine microbiota (69, 70). PEG treatment also induced sustained alterations in the intestinal microbial composition of these mice, via disruption of the mucus barrier, an altered immune response, and increased osmolality (69). In humans, PEG-based bowel cleansing in preparation for sigmoidoscopy results in analogous decreases in microbial diversity and changes in the composition of the mucosa-associated microbiota (71). The effect of chronic laxative use and laxative subtypes with different mechanisms of action on the intestinal environment, including resident microbes, requires further investigation.

We speculated that fluctuating food availability associated with binge eating and dietary restraint might create an environment in the gut that selects for specific bacteria; however, our results suggest that the frequency of binge eating and restraint alone might not be sufficient to predict gut microbial composition. Wearable technologies and mobile applications that measure binge-eating episodes more objectively, accurately, and in real-time can provide more granular information about eating disorder behaviors throughout the day, and can be matched to the timing of stool sample collection (37). Assessment of intestinal transit time—particularly among those who use laxatives—may further enhance our ability to make rigorous inferences about how a particular eating event links with measurements of the gut microbiota given that both food intake and the intestinal microbiota may be under the influence of circadian rhythms and could exhibit diurnal or ultradian fluctuations (72). It is worth noting that our study subsample reported a lower frequency of binge-eating episodes than excluded BEGIN participants, so a larger sample that is more representative of the spectrum of eating disorder pathology may be necessary to detect meaningful associations of binge eating with the intestinal microbiota.

Mechanistic study designs are needed to establish whether a causal relationship between BED or BN behaviors and the intestinal microbiota exists; however, there may also be a bidirectional relationship. For example, gut microbes can synthesize and consume neurotransmitters implicated in both behavioral disorders and in gut motility such as serotonin, γ-Aminobutyric acid (i.e., GABA), and the catecholamines dopamine and epinephrine (73). Large case-control studies with longitudinal designs could generate hypotheses about specific gut microbial genera or microbial metabolic functions that may be implicated in BED or BN etiology, progression, and maintenance, and assess the temporality and durability of our findings. Such well-powered studies could utilize integrative computational techniques to understand the directionality of the relationships amongst multiple variables—thus opening new avenues for therapeutic intervention (74). For example, targeted pre- or probiotics, fecal microbiota transplantation (i.e., the transfer of ‘healthy’ donor feces to an individual with disease) (75), and dietary interventions aimed at ‘enhancing’ the health-promoting genetic potential of the gut microbiota (76) may all be viable complements to current standards of care for BED or BN, pending further research.

This study includes important limitations. Most notably, the type of laxatives used was not available. Whereas some laxative types enhance fluid retention, others decrease net absorption of fluid or alter gut motility (60). Stimulant laxatives can change nerve and muscle function, whereas saline or osmotic laxatives perturb electrolyte and mineral balance (77). The course of action of different types of laxatives can also differ vastly—from hours to days. The differential consequences of laxative subtypes and the related changes in the structure and chemistry of the intestinal environment may lead to the preferential proliferation of specific microbes over others and to enteric dysfunction, particularly if laxative abuse is sustained (60). It may be that laxative type is more predictive of changes in gut microbes than laxative frequency, the latter of which did not predict the composition or diversity of the intestinal microbiota in our study. Eating disorder behaviors themselves (e.g., binge eating, vomiting) could theoretically modify the effect of laxative use on the intestinal microbiota, so future comparisons of the gut microbiome to otherwise healthy participants who use laxatives to achieve bowel regularity without other major medical conditions that could affect the intestinal microbiota may be useful.

Future studies should also systematically collect information about medications used to treat BED and BN. Despite reported side-effects such as irregular bowel movements and weight-gain (78), selective serotonin reuptake inhibitors commonly used in BED and BN treatment are also often prescribed to treat irritable bowel syndrome (79), further complicating the relationship between behavioral features of BED or BN and the intestinal microbiota. Of note, no studies have investigated the effect of lisdexamfetamine on the composition of the gut microbiota, to our knowledge.

Another important limitation is the lack of a quantitative measure of dietary intake, which varies between individuals with and without BED or BN (80) and is an important predictor of intestinal microbial composition and diversity (34). Alongside dietary intake, future studies should measure the fecal metabolome and the host and microbial transcriptome, to estimate the joint influence of diet, laxative use, and the composition and functionality of the intestinal microbiota and its metabolites on functional gastrointestinal disorders, which can exacerbate eating disorder symptoms and are common among individuals with eating disorders who report laxative use (81). Finally, future studies should compare gut microbial composition, diversity, and metabolites between cases with BED or BN and controls to identify microbial signatures of these disorders or commensals that are lacking in sufficient quantities and could therefore be restored with dietary or supplement approaches for eating disorder prevention or treatment.

This study also includes several strengths. This is the first study, to our knowledge, to evaluate the composition of the gut microbiota in people with BN, and to assess relationships of specific eating disorder behaviors in individuals with BED and BN with the composition and diversity of the intestinal microbiota. Our results reveal an important and previously unreported link between laxative use and a global reduction in diversity metrics of the intestinal microbiota as well as in several specific gut microbial taxa—some of which are genera encompassing known beneficial commensals. Furthermore, the directionality of our estimates is remarkably consistent and aligns with the general principle that a “heathy” gut microbiota is characterized by a diverse array of taxa, while gut “dysbiosis” is characterized by a depleted microbial diversity (82). These findings are noteworthy given that laxative abuse is common among individuals with BN (46.5% according to a large, multi-site genetic study) and can lead to serious medical complications, greater eating disorder symptomatology, and poorer treatment outcomes (59). Another strength of our study is the inclusion of males—an underrepresented group in eating disorders research.

A hypothesis that stems from our findings is that the deleterious effects of laxative abuse on gut and systemic health are further exacerbated by reductions in specific intestinal microbes and intestinal microbial diversity. A global depletion of commensal gut microbes could theoretically invite opportunistic infections or reduce the functional potential of gut microbes to produce metabolites that promote intestinal and systemic health, worsening eating disorder phenotypes (83). Ultimately, interventions that promote the maintenance or restoration of a healthy milieu of intestinal microbes and gut mucosal integrity in patients with eating disorders could incentivize and hasten recovery and potentially mitigate some of the challenging co-occurring gastrointestinal disturbances seen in these life-impairing diseases (3, 84).

Abbreviations:

AN

anorexia nervosa

BMI

body mass index

CBT

cognitive behavioral therapy

BED

binge eating disorder

BEGIN

The Binge Eating Genetics Initiative

BN

bulimia nervosa

DADA2

Divisive Amplicon Denoising Algorithm

DSM-5

Diagnostic and Statistical Manual of Mental Disorders, 5^th Edition (DSM-5)

EDE-Qv6

Eating Disorder Examination Questionnaire, version 6

FDR

false discovery rate

IPT

interpersonal psychotherapy

IQR

interquartile range

QIIME2

Quantitative Insights Into Microbial Ecology 2

SSRI

selective serotonin reuptake inhibitor