Preliminary Evidence That Shared Genetic Influences Underlie Comorbidity Between Self‐Reported Eating and Internalizing Disorders and Gastrointestinal Disease in Adult Women and Men

ABSTRACT

Introduction

Accumulating research suggests both eating disorders (EDs) and internalizing disorders (e.g., anxiety, depression) are associated with gastrointestinal disease (e.g., irritable bowel syndrome, inflammatory bowel disease). However, the mechanisms underlying comorbidity with gastrointestinal disease—and whether they may differ for eating and internalizing disorders—remain poorly understood. Addressing these gaps is a critical first step to refining etiologic models of comorbidity and identifying potential targets for intervention.

Method

Participants included female and male twins ages 18–65 from the population‐based MSU Twin Registry (N = 5883). Lifetime history of EDs, internalizing disorders, and gastrointestinal disease was assessed via questionnaire. We first examined whether EDs and internalizing disorders were independently associated with gastrointestinal disease phenotypically. We then used trivariate Cholesky decomposition twin models to investigate whether EDs and internalizing disorders were related to gastrointestinal disease through overlapping or distinct genetic/environmental pathways.

Results

Eating (OR = 2.54, p = 0.009) and internalizing (OR = 2.14, p < 0.001) disorders were independently associated with gastrointestinal disease. Conclusions were unchanged after adjusting for important covariates (e.g., body mass index, age) and did not significantly differ across sex. Twin models suggested genetic influences shared by all three conditions explained their co‐occurrence, with 31% of the variance in EDs and 12% of the variance in gastrointestinal disease attributable to genetic influences shared with internalizing disorders.

Conclusion

Shared genetic mechanisms may contribute to comorbidity between EDs, internalizing disorders, and gastrointestinal disease. Identifying overlapping molecular pathways could potentially lead to novel interventions that simultaneously address all three conditions.

Summary.

• Growing evidence suggests both eating and internalizing disorders (e.g., anxiety, depression) tend to co‐occur with gastrointestinal disease. However, the reasons underlying this comorbidity remain unclear.

• We found that eating and internalizing disorders each independently predicted the likelihood of having a lifetime gastrointestinal disease.

• We also found evidence that eating, internalizing, and gastrointestinal disorders share genetic influences, suggesting overlapping molecular mechanisms may contribute to the development of all three conditions.

Emerging research suggests eating disorders (EDs) and gastrointestinal diseases (e.g., irritable bowel syndrome [IBS], inflammatory bowel disease [IBD], gastroesophageal reflux disease [GERD]) commonly co‐occur. A recent population‐based study of adult twins from our group found that individuals with a lifetime ED had almost 3× the odds of lifetime gastrointestinal disease, with similar associations for women and men and across anorexia nervosa (AN), bulimia nervosa (BN), and binge‐eating disorder (BED) (Pascoe et al. 2024). Simultaneously, individuals receiving treatment for IBD show elevated ED symptoms (e.g., body dissatisfaction; Wabich et al. 2020) and youth with IBD report increased food/eating preoccupations and weight/shape concerns (Nicholas et al. 2007).

However, the degree to which gastrointestinal diseases are linked to EDs specifically above and beyond associations with other mental health conditions remains unclear. Elevated rates of gastrointestinal disease are also found in women and men with internalizing disorders (e.g., depression, anxiety; Fuller‐Thomson, Lateef, and Sulman 2015; Lee et al. 2009; Sanna et al. 2013), which are commonly comorbid with EDs (Garcia et al. 2020; Ulfvebrand et al. 2015). No studies (including Pascoe et al. 2024) have examined associations between EDs and gastrointestinal disease while accounting for comorbid internalizing disorders, which is crucial to determine whether gastrointestinal disease shows specific and independent links with EDs. Analyses including both types of disorders can also illuminate whether they are associated with gastrointestinal disease additively (i.e., each is independently associated with increased risk) or multiplicatively (i.e., EDs are more strongly associated with gastrointestinal disease for people with lifetime internalizing disorders). Understanding additive and multiplicative effects of eating and internalizing disorders can inform screening efforts to identify individuals most likely to have gastrointestinal disease and facilitate prompt integration of comorbidity into treatment.

Another key unanswered question is whether gastrointestinal diseases are related to EDs through mechanisms that overlap with or differ from those that contribute to comorbidity with other mental health concerns. Historically, gastrointestinal disturbances have been understood as complications resulting from disorder‐specific symptoms (e.g., restriction leading to delayed gastric transit, stomachaches resulting from excessive worry; Brosschot and Van Der Doef 2006; Riedlinger et al. 2020), suggesting distinct pathways to comorbidity. While this is likely true to an extent, it is also possible that eating, internalizing, and gastrointestinal disorders share underlying biological mechanisms that increase risk for all three types of conditions. For example, increased inflammation and altered interoceptive processing are biologically‐based traits that have been implicated in EDs as well as internalizing disorders and gastrointestinal disease (Blumberg 2009; Bonaz et al. 2021; Duncan et al. 2017; Howe and Lynch 2022; Khalsa, Berner, and Anderson 2022; Vinni et al. 2023). Moreover, emerging evidence suggests both EDs (Gong et al. 2023; Pascoe et al. 2024) and internalizing disorders (Eijsbouts et al. 2021; Gong et al. 2023; Logue et al. 2011; Shaw et al. 2024) share genetic influences with gastrointestinal disease (i.e., genetic correlations of ~0.30 to 0.55 in prior twin studies). However, it is unknown whether EDs and internalizing disorders share the same genes with gastrointestinal disease. If the same genes that contribute to comorbidity between EDs and gastrointestinal disease also impact risk for internalizing disorders, this would suggest all three types of conditions share underlying molecular mechanisms that could potentially be targeted to alleviate symptoms across multiple domains. Conversely, if EDs and internalizing disorders are related to gastrointestinal disease through partially or fully distinct etiologic pathways, disorder‐specific interventions may be needed.

In the current study, we first examined the extent to which EDs (AN, BN, BED) and internalizing disorders were uniquely associated with gastrointestinal disease at a phenotypic level in a large, population‐based sample of adult female and male twins. We then used twin models to investigate the extent to which EDs and internalizing disorders were associated with gastrointestinal disease through shared versus distinct genetic/environmental pathways. Inclusion of male and female twins allowed us to examine sex differences in effects, which was important given well‐documented sex differences in rates of EDs (Culbert, Sisk, and Klump 2021) and internalizing disorders (Bangasser and Cuarenta 2021) that may contribute to differential associations with gastrointestinal disease. Because both EDs and internalizing disorders share genetic influences with gastrointestinal disease, we hypothesized that at least some genetic influences would be shared across all three types of conditions.

1. Method

1.1. Participants

Participants were 5883 twins aged 18–65 (M _age = 36.57, SD = 8.04) who completed the Michigan Twins Project (MTP) intake between 2010 and 2016. This was the same sample used to examine overlap between EDs and gastrointestinal disease in Pascoe et al. (2024). The wide age range was ideal for capturing lifetime diagnoses, particularly as some gastrointestinal diseases may not onset until midlife (Dorn‐Rasmussen et al. 2023; Halder et al. 2007). The MTP is a population‐based twin registry that serves as a recruitment pool for research through the Michigan State University Twin Registry (MSUTR; Burt and Klump 2013, 2019; Klump and Burt 2006), which recruits twins via birth records. Response rates are similar or better than those of other twin registries (Burt and Klump 2019). Participants were 50.7% female and 49.4% male based on sex recorded on birth certificates (hereafter referred to as sex) and representative of Michigan in regards to race and ethnicity reported via questionnaire based on US Census categories (see Table 1 for demographic information and sample sizes by zygosity).

TABLE 1.

Descriptive statistics for participant demographics and symptoms (N = 5883).

Participant characteristics	Mean (SD) or % of sample (N)
Age	36.57 (8.04) (range = 18.08–65.84)
Sex
Female	50.7% (2980)
Male	49.4% (2903)
Zygosity (N listed as number of pairs)
Complete opposite sex pairs	0.2% (7)
Complete female monozygotic pairs	23.9% (703)
Complete female dizygotic pairs	11.9% (351)
Complete female pairs of unknown zygosity	0.03% (1)
Women without cotwin data (N participants)	14.7% (863)
Complete male monozygotic pairs	18.7% (549)
Complete male dizygotic pairs	8.7% (257)
Complete male pairs of unknown zygosity	0.03% (1)
Men without cotwin data (N participants)	21.8% (1282)
Race and ethnicity
White (non‐hispanic)	90.6% (5328)
Black/African American (non‐hispanic)	5.4% (320)
Hispanic/Latinx	1.7% (102)
Asian American	0.3% (15)
Native American/American Indian	0.2% (14)
More than one race	1.1% (66)
Other/unknown	0.7% (38)
BMI	27.06 (5.69) (range = 16.46–50.97)
Household income	$75,021 (57,466) (range = $0–$300,000+)
Education level
Less than high school	2.9% (170)
High school graduate	19.1% (1117)
Less than 4 years of college	29.3% (1711)
College graduate (4–6 years of college)	35.4% (2071)
Post‐graduate education	13.3% (779)
Reported having an eating disorder (female)	5.7% (166)
AN (female)	3.0% (89)
BN (female)	2.6% (76)
BED (female)	2.2% (64)
Reported having an eating disorder (male)	1.2% (34)
AN (male)	0.6% (16)
BN (male)	0.6% (18)
BED (male)	0.7% (20)
Reported having an internalizing disorder (female)	37.7% (1058)
Depression (female)	28.1% (827)
Separation anxiety disorder (female)	2.7% (79)
Panic disorder (female)	5.2% (152)
Generalized anxiety disorder (female)	12.5% (352)
Social anxiety disorder (female)	7.8% (220)
Obsessive‐compulsive disorder (female)	4.7% (137)
Post‐traumatic stress disorder (female)	4.0% (117)
Reported having an internalizing disorder (male)	23.4% (656)
Depression (male)	17.0% (482)
Separation anxiety disorder (male)	1.4% (40)
Panic disorder (male)	3.1% (90)
Generalized anxiety disorder (male)	6.6% (188)
Social anxiety disorder (male)	5.8% (163)
Obsessive‐compulsive disorder (male)	3.8% (108)
Post‐traumatic stress disorder (male)	2.2% (62)
Reported having a GI disease (female)	5.3% (158)
Reported having a GI disease (male)	4.7% (137)

Note: The number of participants with any eating disorder and any internalizing disorder is smaller than the sum of participants with each individual disorder because some participants had multiple lifetime disorders. Specifically, 59 participants reported multiple lifetime eating disorders (29.5% of participants with eating disorders) and 724 participants reported multiple lifetime internalizing disorders (42.2% of participants with internalizing disorders). Participants were included in analyses for a given diagnosis even if they had other diagnoses from the same category because diagnostic crossover is common in eating disorders and high rates of comorbidity are common among internalizing disorders, raising concern that samples would have been non‐representative if only including participants with a single lifetime eating or internalizing diagnosis. The number of complete opposite‐sex twin pairs was small because adult opposite‐sex twins (i.e., above age 18) were not explicitly recruited by the Michigan Twins Project. We retained these twins and twins without cotwin data in our sample to maximize sample size and include all available participants in phenotypic analyses, as well as maintain consistency with Pascoe et al. (2024). N's may not add up to the total sample size for all variables due to missing values.

Abbreviations: AN = anorexia nervosa; BED = binge‐eating disorder; BMI = body mass index; BN = bulimia nervosa; GI = gastrointestinal; household income = participant and spouse's (if applicable) combined income.

1.2. Measures

1.2.1. Zygosity

Zygosity was determined using a well‐validated physical similarity questionnaire completed by each twin (Lykken et al. 1990). This questionnaire is over 95% accurate in determining zygosity based on DNA and serologic testing (Lykken et al. 1990; Peeters et al. 1998).

1.2.2. ED and Internalizing Diagnoses

During MTP enrollment, participants indicated whether they (1) ever had and (2) had been treated for several mental and physical health conditions, including eating and internalizing disorders. Consistent with Pascoe et al. (2024), we categorized participants based on whether they reported having a diagnosis regardless of whether they sought treatment, as only a minority of individuals with EDs (Hart et al. 2011) and internalizing disorders (Collins et al. 2004) seek treatment. Participants were categorized as having a lifetime ED if they endorsed having AN, BN, or BED. Participants were categorized as having a lifetime internalizing disorder if they endorsed having depression, obsessive‐compulsive disorder (OCD), post‐traumatic stress disorder (PTSD), or a DSM‐5 anxiety disorder (separation anxiety disorder, panic disorder, social anxiety, or generalized anxiety disorder [GAD]). To maximize power, we first combined across EDs (i.e., any/no ED) and internalizing disorders (i.e., any/no internalizing disorder), but individual disorders were also examined when sample sizes allowed. Rates of EDs (women: 5.7%; men: 1.2%) and internalizing disorders (women: 37.7%; men: 23.4%) aligned with other population‐based studies (Hasin et al. 2018; Hay, Girosi, and Mond 2015; Hudson et al. 2007; Kessler et al. 2012). As expected, women had significantly higher rates of both EDs and internalizing disorders than men (ps < 0.001). Rates of internalizing disorders were higher in participants with a lifetime ED relative to those without for both women (74.9% vs. 34.9%) and men (73.5% vs. 22.6%) (ps < 0.001).

1.2.3. Disordered Eating Symptoms

Because this sample lacked a dimensional measure of internalizing symptoms, primary analyses focused on the presence/absence of categorical disorders for both EDs and internalizing disorders. However, a limitation of this approach is that relatively few participants had threshold EDs, reducing power to detect effects. We therefore conducted supplemental analyses with current dimensional ED symptoms measured by the Michigan Twins Project Eating Disorder Survey (MTP‐ED; Mikhail et al. 2021). The MTP‐ED is a nine‐item questionnaire containing items regarding body dissatisfaction, weight preoccupation, and ED behaviors (dieting, purging, binge eating). Each item is rated from 0 (not true) to 2 (certainly true). The MTP‐ED has strong correlations with other ED measures (e.g., r = 0.77 with the Eating Disorder Examination Questionnaire; Mikhail et al. 2021) and distinguishes between men and women with and without lifetime EDs (Mikhail et al. 2023). Internal consistency was excellent for both women (α = 0.86) and men (α = 0.83) and both women (p < 0.001, d = 0.81) and men (p < 0.001, d = 1.26) with self‐reported lifetime EDs had higher MTP‐ED scores in this sample. MTP‐ED scores were prorated if one item was missing and coded as missing if > 1 item was missing.

1.2.4. Gastrointestinal Disease Diagnoses

As reported in Pascoe et al. (2024), the MTP checklist of health conditions did not contain items specifically pertaining to gastrointestinal disease. The presence/absence of gastrointestinal disease was instead coded based on participants' responses to two write‐in items regarding additional chronic illnesses (“any other chronic illness or disability; if ‘Yes’, please describe”) and medication use (“currently taking any prescription medications; if ‘Yes,’ specify what medications and for what conditions”). A participant was coded as having a gastrointestinal disease if they reported (1) having a disease that primarily affects the gastrointestinal tract and/or (2) taking a medication predominantly used to treat a gastrointestinal disease, unless they specified taking it for a non‐gastrointestinal condition. Conditions that may involve gastrointestinal symptoms but do not primarily affect the gastrointestinal tract (e.g., nausea associated with migraines) were not included. The first and second authors (LAP and MEM) developed a coding guide for inclusion/exclusion of gastrointestinal diseases (see Supporting Information), then independently coded the presence/absence of gastrointestinal disease for each participant. Interrater reliability was excellent (κ = 0.97; Pascoe et al. 2024) and all discrepancies were resolved by consensus of MEM and LAP.

Overall, 295 (5.0%) participants reported a gastrointestinal disease, consistent with rates in other population‐based samples (Canavan, West, and Card 2014; El‐Serag et al. 2014). Gastrointestinal diseases included GERD/heartburn (n = 172; 58.3% of participants with gastrointestinal diseases), IBS (n = 23; 7.8%), Crohn's disease (n = 18; 6.1%), ulcerative colitis/colitis (n = 14; 4.7%), gastrointestinal ulcers (n = 8; 2.7%), diverticulitis (n = 6; 2.0%), gastritis (n = 4; 1.4%), celiac disease (n = 3; 1.0%), gastroparesis (n = 1; 0.3%), colon cancer (n = 1; 0.3% this participant also endorsed using Prilosec for gastroesophageal reflux), or another condition clearly linked to gastrointestinal health (e.g., “chronic stomach problems,” “gastrointestinal disease”; n = 9; 3.0%). Participants who reported taking a medication primarily used to treat gastrointestinal disease (e.g., omeprazole, Nexium, Asacol; n = 46; 15.6%) but did not name a specific diagnosis were also included in the group with gastrointestinal diseases. A small number of participants provided responses that were difficult to definitively code (e.g., “chronic pain in my lower right side,” “autoimmune disorder” without specification; n = 12, 0.2%) and were excluded. The presence of gastrointestinal disease was coded dichotomously (any/no gastrointestinal disease) to maximize power due to small sample sizes for individual conditions. This approach was supported by analyses from Pascoe et al. (2024) showing similar associations between different gastrointestinal diseases and EDs.

1.2.5. Covariates

1.2.5.1. Body Mass Index (BMI)

People in larger bodies are more likely to experience disordered eating, in part due to weight stigma (Puhl and Suh 2015). Research also suggests elevated rates of internalizing disorders among people who are underweight and at higher weights (DeJesus et al. 2016; Zhao et al. 2009), which may reflect the symptomatology of some internalizing disorders (e.g., appetite changes in depression). We conducted phenotypic analyses with and without covarying BMI calculated from self‐reported height and weight to directly assess its impact on associations. To remove extreme values likely representing errors in self‐reported height and weight, we set BMI values above the 99.5th percentile or below the 0.5th percentile within the sample to missing (note that conclusions were unchanged if all BMI values were included; see Table S1).

1.2.5.2. Demographic Covariates

Demographic covariates included sex, age, and household income (the sum of the participant's income and their spouse's income, if applicable). We did not include racial identity as a covariate in primary analyses to avoid estimation errors resulting from the small number of participants with some racial identities and conceptual issues associated with grouping participants with distinct identities together. However, conclusions were unchanged if racial identity was included as a dichotomized covariate (White participants compared to participants of color; see Table S2).

1.3. Statistical Analyses

1.3.1. Phenotypic Analyses

Phenotypic analyses used logistic multilevel models with maximum likelihood estimation and a random intercept at the family level to account for clustering of twins within families. Analyses predicted odds of gastrointestinal disease from presence of a lifetime ED, lifetime internalizing disorder, and covariates. Continuous variables (age, income, BMI) were z‐scored to facilitate interpretability. As in Pascoe et al. (2024), sex was initially included as a covariate, then examined as a moderator of associations with gastrointestinal disease. We also used moderation models to investigate whether EDs were more strongly associated with gastrointestinal disease for participants with lifetime internalizing disorders.

1.3.2. Twin Analyses

Trivariate Cholesky decomposition twin analyses (Neale and Cardon 1992) were used to examine whether EDs were related to gastrointestinal disease through genetic/environmental effects partially or fully overlapping with those on internalizing disorders. Broadly, Cholesky twin models estimate additive genetic effects (A; genetic effects that sum across genes), shared environmental effects (C; environmental effects that increase cotwin similarity, such as family socioeconomic background), and nonshared environmental effects (E; environmental effects that decrease cotwin similarity, such as individual bullying experiences) on a given trait and the extent to which these are shared across traits. As in Pascoe et al. (2024), we excluded the small number of opposite‐sex twin pairs (7 pairs; 0.2%) from these analyses to prevent parameter estimate biases related to possible qualitative sex differences in etiologic effects. Twins without cotwin data were included because they aid in estimation of the total variance for each trait (Neale and Cardon 1992); however, conclusions were unchanged if they were excluded (see Table S3).

Consistent with Pascoe et al. (2024), covariates were not included in twin models due to computational challenges related to use of dichotomous outcomes. However, we did examine possible sex differences in patterns of genetic/environmental effects. All parameters could be constrained to equality across sex without significantly worsening model fit (ps ≥ 0.722 for primary analyses; see Table S4), suggesting a similar pattern of genetic/environmental effects in women and men. Nevertheless, full model estimates in women and men separately are included in Table S5.

Analyses were conducted in Mplus version 8.6 (Muthén and Muthén 1998–2021) using WLSMV estimation (Bandalos 2014) based on liability threshold models from Prescott (2004). The full model is depicted in Figure 1. A₁, C₁, and E₁ in Figure 1 represent genetic, shared environmental, and nonshared environmental effects on internalizing disorders. Arrows from A₁ and E₁ to gastrointestinal disease and EDs indicate genetic and nonshared environmental effects on internalizing disorders that also influence gastrointestinal disease and EDs, respectively. The model also allows for genetic (A₂) and nonshared environmental (E₂) effects on gastrointestinal disease that are not shared with internalizing disorders. The arrows from A₂ and E₂ to EDs indicate genetic and nonshared environmental effects on gastrointestinal disease independent of internalizing disorders that also influence EDs. Finally, A₃ and E₃ in the model represent genetic and nonshared environmental effects unique to EDs. Although not explicitly specified in the model, correlations between genetic/environmental influences on each trait can be derived from model parameters. Note that we did not include shared environmental effects on EDs or gastrointestinal disease because Pascoe et al. (2024) indicated these were minimal and could be constrained to zero. However, we included shared environmental effects for internalizing disorders because we had not previously examined whether these could be constrained to zero.

FIGURE 1.

Full and best‐fitting trivariate etiologic models for internalizing disorders, gastrointestinal disease, and eating disorders. GI = gastrointestinal; A₁ = additive genetic effects on internalizing disorders; C₁ = shared environmental effects on internalizing disorders; E₁ = nonshared environmental effects on internalizing disorders; A₂ = additive genetic effects on gastrointestinal disease not shared with internalizing disorders; E₂ = nonshared environmental effects on gastrointestinal disease not shared with internalizing disorders; A₃ = additive genetic effects on eating disorders not shared with internalizing disorders or gastrointestinal disease; E₃ = nonshared environmental effects on eating disorders not shared with internalizing disorders or gastrointestinal disease. Path estimates are standardized with 95% confidence intervals in parentheses. Significant parameters are bolded and non‐significant paths are denoted with dotted lines.

After fitting the full model, we fit nested submodels that constrained parameters to zero based on full model parameter estimates to identify a best‐fitting model. Because twin models with categorical outcomes in Mplus do not provide some fit indices traditionally used to compare models, such as AIC/BIC, we primarily relied on chi‐square difference tests optimized for WLSMV estimation (implemented with the DIFFTEST option) to choose a best‐fitting model. We also report TLI, RMSEA, and SRMR as indicators of absolute model fit. Recommendations in the literature suggest adequate model fit if RMSEA < 0.08 (Browne and Cudeck 1993) and TLI ≥ 0.95 or SRMR ≤ 0.09 (Hu and Bentler 1999). Confidence intervals were calculated using percentile bootstraps with 1000 random samples with replacement.

2. Results

2.1. Phenotypic Associations Between EDs and Internalizing Disorders and Gastrointestinal Disease

Lifetime EDs (OR = 2.54, p = 0.009) and internalizing disorders (OR = 2.14, p < 0.001) were significantly and independently associated with greater odds of lifetime gastrointestinal disease (see Table 2). Similar patterns were observed for depression (ED OR = 2.69, p = 0.007; depression OR = 1.68, p = 0.005), any DSM‐5 anxiety disorder (ED OR = 2.75, p = 0.005; anxiety OR = 2.21, p < 0.001), and PTSD (ED OR = 2.72, p = 0.007; PTSD OR = 3.67, p < 0.001). Interestingly, in models of OCD, only lifetime EDs were significantly associated with gastrointestinal disease (ED OR = 3.15, p = 0.002; OCD OR = 1.34, p = 0.433). Conclusions were generally unchanged when covarying BMI, although the association between EDs and gastrointestinal disease was only present at a trend level in the model with PTSD that covaried BMI (see Table 2).

TABLE 2.

Eating and internalizing disorders as phenotypic predictors of gastrointestinal disease.

Any internalizing disorder
Predictor	Odds ratio	SE	p	95% Confidence interval
BMI excluded
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Internalizing disorder	2.14	0.38	< 0.001	[1.52, 3.03]
ED	2.54	0.91	0.009	[1.26, 5.11]
Age	1.66	0.17	< 0.001	[1.35, 2.04]
Sex	0.91	0.16	0.585	[0.64, 1.29]
Income	1.04	0.09	0.651	[0.87, 1.24]
BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Internalizing disorder	2.00	0.37	< 0.001	[1.39, 2.87]
ED	2.19	0.87	0.047	[1.01, 4.76]
Age	1.64	0.18	< 0.001	[1.31, 2.04]
Sex	0.92	0.17	0.638	[0.64, 1.32]
Income	1.07	0.10	0.446	[0.89, 1.29]
BMI	1.23	0.11	0.015	[1.04, 1.46]
Interactions with sex—BMI excluded
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Internalizing disorder	2.13	0.50	0.001	[1.35, 3.36]
Internalizing disorder × sex	1.02	0.35	0.960	[0.52, 2.01]
ED	2.61	1.02	0.014	[1.21, 5.63]
ED × sex	0.84	0.80	0.851	[0.13, 5.47]
Sex	0.91	0.20	0.662	[0.59, 1.41]
Age	1.66	0.17	< 0.001	[1.35, 2.04]
Income	1.04	0.09	0.659	[0.87, 1.24]
Interactions with sex—BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Internalizing disorder	1.86	0.47	0.014	[1.13, 3.04]
Internalizing disorder × sex	1.18	0.43	0.654	[0.58, 2.39]
ED	2.51	1.08	0.033	[1.08, 5.85]
ED × sex	0.46	0.52	0.492	[0.05, 4.21]
Sex	0.88	0.21	0.580	[0.56, 1.39]
Age	1.64	0.18	< 0.001	[1.31, 2.04]
Income	1.07	0.10	0.462	[0.89, 1.29]
BMI	1.24	0.11	0.014	[1.04, 1.47]
Interaction between EDs and internalizing disorders—BMI excluded
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Internalizing disorder	2.10	0.38	< 0.001	[1.47, 2.98]
ED	1.65	1.33	0.536	[0.34, 8.01]
Internalizing disorder × ED	1.72	1.54	0.544	[0.30, 9.95]
Age	1.67	0.18	< 0.001	[1.36, 2.05]
Sex	0.90	0.16	0.572	[0.64, 1.28]
Income	1.04	0.09	0.638	[0.88, 1.24]
Interaction between EDs and internalizing disorders—BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Internalizing disorder	1.99	0.37	< 0.001	[1.37, 2.87]
ED	1.93	1.62	0.432	[0.37, 9.98]
Internalizing disorder × ED	1.18	1.11	0.863	[0.18, 7.51]
Age	1.64	0.18	< 0.001	[1.31, 2.04]
Sex	0.92	0.17	0.634	[0.64, 1.32]
Income	1.07	0.10	0.444	[0.89, 1.29]
BMI	1.23	0.11	0.015	[1.04, 1.46]
Depression
BMI excluded
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Depression	1.68	0.31	0.005	[1.17, 2.43]
ED	2.69	0.98	0.007	[1.32, 5.48]
Age	1.73	0.18	< 0.001	[1.41, 2.13]
Sex	0.90	0.16	0.557	[0.64, 1.28]
Income	1.04	0.09	0.696	[0.87, 1.23]
BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Depression	1.51	0.30	0.038	[1.02, 2.23]
ED	2.35	0.95	0.035	[1.06, 5.18]
Age	1.71	0.19	< 0.001	[1.37, 2.14]
Sex	0.90	0.17	0.592	[0.63, 1.30]
Income	1.07	0.10	0.481	[0.89, 1.28]
BMI	1.24	0.11	0.013	[1.05, 1.47]
Interactions with sex—BMI excluded
Intercept	0.01	0.003	< 0.001	[0.005, 0.02]
Depression	1.75	0.42	0.019	[1.10, 2.81]
Depression × sex	0.90	0.34	0.785	[0.43, 1.90]
ED	2.70	1.07	0.012	[1.24, 5.88]
ED × sex	0.92	0.89	0.936	[0.14, 6.15]
Sex	0.93	0.19	0.723	[0.62, 1.39]
Age	1.73	0.18	< 0.001	[1.41, 2.13]
Income	1.03	0.09	0.713	[0.87, 1.23]
Interactions with sex—BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Depression	1.49	0.39	0.130	[0.89, 2.48]
Depression × sex	1.05	0.42	0.903	[0.48, 2.29]
ED	2.61	1.15	0.029	[1.10, 6.20]
ED × sex	0.51	0.59	0.560	[0.06, 4.81]
Sex	0.91	0.20	0.660	[0.60, 1.39]
Age	1.71	0.19	< 0.001	[1.37, 2.14]
Income	1.07	0.10	0.498	[0.89, 1.28]
BMI	1.24	0.11	0.012	[1.05, 1.47]
Interaction between EDs and depression—BMI excluded
Intercept	0.01	0.003	< 0.001	[0.005, 0.02]
Depression	1.57	0.30	0.020	[1.07, 2.29]
ED	1.05	0.82	0.950	[0.23, 4.81]
Depression × ED	3.58	3.15	0.147	[0.64, 20.06]
Age	1.74	0.18	< 0.001	[1.42, 2.14]
Sex	0.89	0.16	0.519	[0.63, 1.26]
Income	1.04	0.09	0.666	[0.87, 1.23]
Interaction between EDs and depression—BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Depression	1.44	0.30	0.074	[0.96, 2.16]
ED	1.21	0.98	0.811	[0.25, 5.91]
Depression × ED	2.51	2.34	0.324	[0.40, 15.57]
Age	1.72	0.20	< 0.001	[1.38, 2.15]
Sex	0.90	0.17	0.564	[0.62, 1.29]
Income	1.07	0.10	0.474	[0.89, 1.29]
BMI	1.24	0.11	0.013	[1.05, 1.47]
Any DSM‐5 Anxiety disorder
BMI excluded
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Anxiety disorder	2.21	0.46	< 0.001	[1.47, 3.32]
ED	2.75	1.00	0.005	[1.35, 5.61]
Age	1.70	0.18	< 0.001	[1.38, 2.09]
Sex	0.90	0.16	0.557	[0.64, 1.28]
Income	1.03	0.09	0.710	[0.87, 1.23]
BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Anxiety disorder	1.97	0.44	0.002	[1.27, 3.04]
ED	2.44	0.98	0.027	[1.11, 5.36]
Age	1.66	0.19	< 0.001	[1.32, 2.07]
Sex	0.91	0.17	0.592	[0.63, 1.30]
Income	1.07	0.10	0.460	[0.89, 1.29]
BMI	1.27	0.11	0.006	[1.07, 1.51]
Interactions with sex—BMI excluded
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Anxiety disorder	2.41	0.64	0.001	[1.43, 4.07]
Anxiety disorder × sex	0.80	0.34	0.599	[0.35, 1.82]
ED	2.80	1.11	0.010	[1.28, 6.10]
ED × sex	0.85	0.82	0.865	[0.13, 5.62]
Sex	0.95	0.19	0.785	[0.64, 1.39]
Age	1.70	0.18	< 0.001	[1.38, 2.09]
Income	1.03	0.09	0.718	[0.87, 1.23]
Interactions with sex—BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Anxiety disorder	2.03	0.59	0.015	[1.15, 3.59]
Anxiety disorder × sex	0.94	0.42	0.894	[0.40, 2.24]
ED	2.74	1.20	0.021	[1.16, 6.49]
ED × sex	0.47	0.53	0.503	[0.05, 4.28]
Sex	0.94	0.19	0.746	[0.62, 1.40]
Age	1.66	0.19	< 0.001	[1.33, 2.07]
Income	1.07	0.10	0.487	[0.89, 1.29]
BMI	1.28	0.11	0.005	[1.07, 1.52]
Interaction between EDs and Anxiety—BMI excluded
Intercept	0.01	0.003	< 0.001	[0.005, 0.02]
Anxiety	2.13	0.47	0.001	[1.39, 3.27]
ED	2.25	1.18	0.122	[0.80, 6.28]
Anxiety × ED	1.48	1.06	0.584	[0.36, 6.02]
Age	1.70	0.18	< 0.001	[1.38, 2.09]
Sex	0.90	0.16	0.545	[0.63, 1.27]
Income	1.03	0.09	0.703	[0.87, 1.23]
Interaction between EDs and Anxiety—BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
Anxiety	1.93	0.45	0.004	[1.23, 3.04]
ED	2.19	1.25	0.169	[0.72, 6.70]
Anxiety × ED	1.24	0.98	0.789	[0.26, 5.89]
Age	1.66	0.19	< 0.001	[1.33, 2.07]
Sex	0.90	0.17	0.586	[0.63, 1.30]
Income	1.07	0.10	0.460	[0.89, 1.29]
BMI	1.27	0.11	0.006	[1.07, 1.51]
Obsessive‐compulsive disorder (OCD)
BMI excluded
Intercept	0.01	0.003	< 0.001	[0.005, 0.02]
OCD	1.34	0.50	0.433	[0.64, 2.79]
ED	3.15	1.16	0.002	[1.54, 6.47]
Age	1.75	0.18	< 0.001	[1.42, 2.14]
Sex	0.87	0.15	0.444	[0.62, 1.23]
Income	1.02	0.09	0.859	[0.85, 1.21]
BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
OCD	1.15	0.48	0.732	[0.51, 2.59]
ED	2.79	1.14	0.012	[1.25, 6.21]
Age	1.70	0.19	< 0.001	[1.36, 2.13]
Sex	0.91	0.17	0.594	[0.63, 1.30]
Income	1.07	0.10	0.490	[0.89, 1.28]
BMI	1.26	0.11	0.010	[1.06, 1.49]
Interactions with sex—BMI excluded
Intercept	0.01	0.003	< 0.001	[0.01, 0.02]
OCD	1.35	0.66	0.538	[0.52, 3.51]
OCD × sex	0.99	0.75	0.985	[0.22, 4.37]
ED	3.20	1.29	0.004	[1.46, 7.05]
ED × sex	0.90	0.87	0.917	[0.14, 6.01]
Sex	0.88	0.16	0.471	[0.62, 1.25]
Age	1.75	0.18	< 0.001	[1.42, 2.14]
Income	1.02	0.09	0.864	[0.85, 1.21]
Interactions with sex—BMI included
Intercept	0.01	0.004	< 0.001	[0.005, 0.02]
OCD	0.95	0.55	0.930	[0.30, 2.96]
OCD × sex	1.53	1.27	0.606	[0.30, 7.75]
ED	3.20	1.43	0.009	[1.33, 7.70]
ED × sex	0.47	0.54	0.513	[0.05, 4.48]
Sex	0.91	0.17	0.606	[0.62, 1.32]
Age	1.71	0.19	< 0.001	[1.37, 2.13]
Income	1.06	0.10	0.520	[0.88, 1.28]
BMI	1.25	0.11	0.009	[1.06, 1.50]
Interaction between EDs and OCD—BMI excluded
Intercept	0.01	0.003	< 0.001	[0.01, 0.02]
OCD	1.39	0.59	0.436	[0.60, 3.21]
ED	3.27	1.34	0.004	[1.46, 7.31]
OCD × ED	0.85	0.75	0.854	[0.15, 4.79]
Age	1.75	0.18	< 0.001	[1.42, 2.15]
Sex	0.87	0.15	0.445	[0.62, 1.23]
Income	1.02	0.09	0.860	[0.85, 1.21]
Interaction between EDs and OCD—BMI included
Intercept	0.01	0.004	< 0.001	[0.005, 0.02]
OCD	1.16	0.54	0.757	[0.46, 2.88]
ED	2.80	1.28	0.025	[1.14, 6.85]
OCD × ED	0.99	0.98	0.988	[0.14, 6.97]
Age	1.70	0.19	< 0.001	[1.36, 2.13]
Sex	0.91	0.17	0.594	[0.63, 1.30]
Income	1.07	0.10	0.490	[0.89, 1.28]
BMI	1.26	0.11	0.010	[1.06, 1.49]
Post‐traumatic stress disorder (PTSD)
BMI excluded
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
PTSD	3.67	1.37	< 0.001	[1.77, 7.62]
ED	2.72	1.00	0.007	[1.32, 5.58]
Age	1.71	0.18	< 0.001	[1.39, 2.11]
Sex	0.86	0.15	0.409	[0.61, 1.22]
Income	1.05	0.09	0.591	[0.88, 1.25]
BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
PTSD	3.91	1.60	0.001	[1.76, 8.71]
ED	2.21	0.92	0.056	[0.98, 4.98]
Age	1.69	0.19	< 0.001	[1.35, 2.11]
Sex	0.88	0.17	0.490	[0.61, 1.27]
Income	1.10	0.11	0.331	[0.91, 1.32]
BMI	1.25	0.11	0.013	[1.05, 1.49]
Interactions with sex—BMI excluded
Intercept	0.01	0.003	< 0.001	[0.005, 0.02]
PTSD	2.47	1.16	0.056	[0.98, 6.22]
PTSD × sex	2.92	2.16	0.150	[0.68, 12.50]
ED	2.95	1.19	0.007	[1.34, 6.49]
ED × sex	0.74	0.73	0.757	[0.11, 5.12]
Sex	0.82	0.15	0.289	[0.57, 1.18]
Age	1.72	0.18	< 0.001	[1.40, 2.12]
Income	1.05	0.09	0.612	[0.88, 1.25]
Interactions with sex—BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
PTSD	2.29	1.23	0.125	[0.80, 6.59]
PTSD × sex	3.67	2.95	0.106	[0.76, 17.70]
ED	2.72	1.23	0.027	[1.12, 6.59]
ED × sex	0.40	0.44	0.403	[0.04, 3.82]
Sex	0.84	0.16	0.374	[0.57, 1.23]
Age	1.70	0.20	< 0.001	[1.36, 2.14]
Income	1.09	0.11	0.362	[0.90, 1.32]
BMI	1.25	0.11	0.014	[1.05, 1.49]
Interaction between EDs and PTSD—BMI excluded
Intercept	0.01	0.003	< 0.001	[0.005, 0.02]
PTSD	3.93	1.62	0.001	[1.76, 8.80]
ED	2.90	1.17	0.008	[1.31, 6.40]
PTSD × ED	0.71	0.64	0.704	[0.12, 4.11]
Age	1.71	0.18	< 0.001	[1.39, 2.11]
Sex	0.86	0.15	0.414	[0.61, 1.23]
Income	1.05	0.09	0.597	[0.88, 1.25]
Interaction between EDs and PTSD—BMI included
Intercept	0.01	0.003	< 0.001	[0.004, 0.02]
PTSD	4.08	1.83	0.002	[1.69, 9.80]
ED	2.32	1.08	0.071	[0.93, 5.76]
PTSD × ED	0.80	0.80	0.824	[0.11, 5.63]
Age	1.69	0.19	< 0.001	[1.35, 2.11]
Sex	0.88	0.17	0.493	[0.61, 1.27]
Income	1.10	0.11	0.333	[0.91, 1.32]
BMI	1.25	0.11	0.013	[1.05, 1.49]

Note: Reference group for sex is female. Race and ethnicity were not included in models to avoid estimation difficulties resulting from the small number of participants in some racial/ethnic groups; however, the overall pattern of results was identical if race/ethnicity was included as a dichotomous covariate (i.e., comparing White participants to participants of color). The number of participants with each condition examined as a predictor was: any ED = 200, any internalizing disorder = 1714, depression = 1309, any DSM‐5 anxiety disorder = 862, OCD = 245, and PTSD = 179. Effects significant at p < 0.05 are bolded.

Abbreviations: Any DSM‐5 anxiety disorder = separation anxiety disorder, social anxiety disorder, generalized anxiety disorder, and/or panic disorder; BMI = body mass index; ED = eating disorder.

Consistent with Pascoe et al. (2024), BED showed the strongest association with gastrointestinal disease among specific ED diagnoses when accounting for internalizing disorders (see Table S6). However, caution is warranted in interpreting analyses of specific EDs because relatively few participants had each diagnosis. Dimensional ED symptoms showed similar, albeit slightly smaller, associations with gastrointestinal disease compared to threshold EDs (see Table S7), which may reflect lower severity associated with dimensional symptoms or the fact that dimensional symptoms were current while diagnoses were lifetime. Associations did not significantly differ across sex and there were no significant interactions between threshold EDs or ED symptoms and internalizing disorders in any models (all ps > 0.05).

2.2. Genetic/Environmental Effects Underlying EDs, Internalizing Disorders, and Gastrointestinal Disease

As hypothesized, the full model suggested that genetic effects on internalizing disorders also contribute to gastrointestinal disease and EDs, as indicated by significant paths from A₁ to EDs and gastrointestinal disease in Figure 1. However, the full model also suggested minimal genetic effects uniquely shared between gastrointestinal disease and EDs after accounting for genetic effects on internalizing disorders, as indicated by a non‐significant path from A₂ to EDs in Figure 1. In other words, all genetic influences shared between EDs and gastrointestinal disease also appeared to be shared with internalizing disorders.

To confirm these impressions with model testing, we constrained genetic/environmental effects shared between gastrointestinal disease and EDs independent of internalizing disorders to zero. We also constrained nonshared environmental effects common to internalizing disorders and gastrointestinal disease to zero given the small effect size observed in the full model. The model with these paths constrained had a non‐significant change in chi‐square and improvements in TLI, RMSEA, and SRMR relative to the full model (see Table 3). It also retained genetic and nonshared environmental effects unique to EDs and gastrointestinal disease. Indeed, although 31% of the variance in EDs and 12% of the variance in gastrointestinal disease was attributable to genetic effects shared with internalizing disorders (with genetic correlations of 0.72 and 0.43, respectively), 28% of the variance in EDs and 52% of the variance in gastrointestinal disease was attributable to unique genetic effects, and 33% of the variance in EDs and 36% of the variance in gastrointestinal disease was attributable to unique environmental effects.

TABLE 3.

Model fit for alternative models of etiologic relationships between eating and internalizing disorders and gastrointestinal disease.

Model	Δχ2 (df)	P	TLI	RMSEA	SRMR
Any internalizing disorder, gastrointestinal disease (GI), and eating disorders (ED)
Full model	—	—	0.996	0.007 [0.000, 0.019]	0.074
Constrain nonshared environmental effects on GI attributable to internalizing	0.829 (1)	0.363	0.997	0.006 [0.000, 0.019]	0.073
Constrain nonshared environmental effects on GI attributable to internalizing and nonshared environmental effects on ED attributable to GI	0.886 (2)	0.642	0.998	0.005 [0.000, 0.018]	0.073
Constrain nonshared environmental effects on GI attributable to internalizing and genetic effects on ED attributable to GI	0.772 (2)	0.680	0.998	0.005 [0.000, 0.018]	0.073
Constrain nonshared environmental effects on GI attributable to internalizing, genetic and nonshared environmental effects on ED attributable to GI	0.838 (3)	0.840	0.999	0.004 [0.000, 0.017]	0.073
Depression, GI, and ED
Full model	—	—	0.999	0.002 [0.000, 0.017]	0.071
Constrain nonshared environmental effects on GI attributable to depression	0.647 (1)	0.421	1.000	0.001 [0.000, 0.017]	0.072
Constrain nonshared environmental effects on GI attributable to depression and nonshared environmental effects on ED attributable to GI	0.692 (2)	0.708	1.000	0.000 [0.000, 0.016]	0.072
Constrain nonshared environmental effects on GI attributable to depression and genetic effects on ED attributable to GI	0.579 (2)	0.749	1.000	0.000 [0.000, 0.016]	0.072
Constrain nonshared environmental effects on GI attributable to depression, genetic and nonshared environmental effects on ED attributable to GI	0.651 (3)	0.885	1.000	0.000 [0.000, 0.016]	0.072
DSM‐5 Anxiety disorders, GI, and ED
Full model	—	—	1.000	0.000 [0.000, 0.010]	0.062
Constrain nonshared environmental effects on ED attributable to GI	0.429 (1)	0.512	1.000	0.000 [0.000, 0.009]	0.062
Constrain nonshared environmental effects on ED attributable to GI and genetic effects on GI attributable to anxiety	2.634 (2)	0.268	1.000	0.000 [0.000, 0.011]	0.065
Constrain nonshared environmental effects on ED attributable to GI and nonshared environmental effects on GI attributable to anxiety	2.948 (2)	0.229	1.000	0.000 [0.000, 0.011]	0.064
Constrain nonshared environmental effects on ED attributable to GI, genetic and nonshared environmental effects on GI attributable to anxiety	23.775 (3)	< 0.001	0.971	0.015 [0.001, 0.024]	0.076
Constrain nonshared environmental and genetic effects on ED attributable to GI	1.365 (2)	0.505	1.000	0.000 [0.000, 0.009]	0.063
Constrain nonshared environmental and genetic effects on ED attributable to GI, genetic effects on GI attributable to anxiety	9.462 (3)	0.024	1.000	0.000 [0.000, 0.016]	0.079
Constrain nonshared environmental and genetic effects on ED attributable to GI, nonshared environmental effects on GI attributable to anxiety	3.032 (3)	0.387	1.000	0.000 [0.000, 0.010]	0.064

Note: Separate twin models were not conducted for PTSD, OCD, or individual eating disorder diagnoses due to small sample sizes. Dashes indicate a parameter is not applicable. 95% confidence interval for RMSEA is presented in brackets. The best‐fitting model description is bolded.

Abbreviations: Δχ² = change in chi‐square; Any DSM‐5 anxiety disorder = separation anxiety disorder, social anxiety disorder, generalized anxiety disorder, and/or panic disorder; df = degrees of freedom; ED = eating disorder; GI = gastrointestinal disease; Internalizing = internalizing disorder; RMSEA = root mean square error of approximation; SRMR = standardized root mean squared residual; TLI = Tucker–Lewis index.

Results were very similar when depression or DSM‐5 anxiety disorder diagnoses were modeled rather than all internalizing diagnoses, with all genetic/environmental effects shared between EDs and gastrointestinal disease also shared with internalizing disorders (see Table 3 and Figures 2, 3). The same was true when dimensional ED symptoms were modeled rather than threshold EDs (see Table S8 and Figure S1).

FIGURE 2.

Full and best‐fitting trivariate etiologic models for depression, gastrointestinal disease, and eating disorders. GI = gastrointestinal; A₁ = additive genetic effects on depression; C₁ = shared environmental effects on depression; E₁ = nonshared environmental effects on depression; A₂ = additive genetic effects on gastrointestinal disease not shared with depression; E₂ = nonshared environmental effects on gastrointestinal disease not shared with depression; A₃ = additive genetic effects on eating disorders not shared with depression or gastrointestinal disease; E₃ = nonshared environmental effects on eating disorders not shared with depression or gastrointestinal disease. Path estimates are standardized with 95% confidence intervals in parentheses. Significant parameters are bolded and non‐significant paths are denoted with dotted lines.

FIGURE 3.

Full and best‐fitting trivariate etiologic models for DSM‐5 anxiety disorders, gastrointestinal disease, and eating disorders. GI = gastrointestinal; A₁ = additive genetic effects on anxiety disorders; C₁ = shared environmental effects on anxiety disorders; E₁ = nonshared environmental effects on anxiety disorders; A₂ = additive genetic effects on gastrointestinal disease not shared with anxiety disorders; E₂ = nonshared environmental effects on gastrointestinal disease not shared with anxiety disorders; A₃ = additive genetic effects on eating disorders not shared with anxiety disorders or gastrointestinal disease; E₃ = nonshared environmental effects on eating disorders not shared with anxiety disorders or gastrointestinal disease. Path estimates are standardized with 95% confidence intervals in parentheses. Significant parameters are bolded. Significant parameters are bolded and non‐significant paths are denoted with dotted lines.

3. Discussion

This was the first study to examine the extent to which eating and internalizing disorders are independently associated with gastrointestinal disease and investigate whether overlapping genetic influences on EDs and gastrointestinal disease identified in prior work (Pascoe et al. 2024) are shared with internalizing disorders. We found that lifetime eating and internalizing disorders were independently (but not multiplicatively) associated with lifetime gastrointestinal disease when included together in phenotypic models. Using twin models, we found evidence that shared genetic factors contribute to EDs, gastrointestinal disease, and internalizing disorders. Results were consistent in women and men, suggesting similar associations and underlying etiology across sex. Overall, findings substantially extend our understanding of shared risk between gastrointestinal disease and mental health concerns.

Given the additive effects observed in this study, people with a history of both an ED and an internalizing disorder may be at especially high risk for gastrointestinal disease and may benefit from targeted screening. Although prior research has not examined eating and internalizing disorders in relation to gastrointestinal disease specifically, results are consistent with evidence that risk for chronic physical conditions (e.g., cancer, chronic lung disease) increases with total number of lifetime psychiatric diagnoses (Scott et al. 2016). People with comorbid mental and physical health conditions may in turn be at risk for poorer outcomes, including reduced functionality and quality of life (Baumeister et al. 2011; Cavalcanti‐Ribeiro et al. 2012) and suicide (Qin et al. 2014). This underscores the importance of early identification of comorbidity to facilitate timely intervention and prevent negative sequelae.

Twin analyses suggested the co‐occurrence of EDs, internalizing disorders, and gastrointestinal disease was best explained by overlapping genetic influences shared across all three types of diagnoses. At first glance, it may seem counterintuitive that EDs and internalizing disorders were associated with gastrointestinal disease independently at a phenotypic level but through a common genetic pathway at an etiologic level. One possible explanation is that the number of disorders a person experiences phenotypically reflects the degree to which they have latent genetic factors that increase risk for all three types of conditions. In other words, as genetic risk through the shared pathway increases (which can rudimentarily be conceptualized as a person's proportion of “risky” alleles), the likelihood for meeting the diagnostic threshold for each condition—and thus of having multiple diagnoses—may also increase. This theoretical model mirrors the hypothesized “p‐factor” that confers risk for multiple psychiatric conditions (Caspi et al. 2024) in part through shared genetic influences (Selzam et al. 2018). An analogous, genetically based “gut‐brain” factor may contribute to the phenotypic expression of EDs, internalizing disorders, and gastrointestinal disease, consistent with the emerging concept of EDs as “neuro‐gastro‐metabolic” disorders (Bulik, Carroll, and Mehler 2021).

An important avenue for future research is to identify the specific genes associated with shared risk for EDs, internalizing disorders, and gastrointestinal disease. One possible shared pathway highlighted in our previous paper on EDs and gastrointestinal disease is genetic factors underlying inflammatory processes (Pascoe et al. 2024). Increased inflammation is common in gastrointestinal disease (Kandulski and Malfertheiner 2012; Xavier and Podolsky 2007) and elevated inflammatory markers have been found in people with EDs (Dalton et al. 2018; MacDowell et al. 2013) and internalizing disorders (Osimo et al. 2020; Passos et al. 2015). Autoimmune disorders are also more common in first‐degree relatives of people with EDs (Zerwas et al. 2017) and internalizing disorders (Nevriana et al. 2022; Zeng et al. 2023), suggesting a potential genetic contribution to increased inflammation. A second possible mechanism is genetic influences on interoceptive processing (i.e., processing of internal bodily sensations). Altered interoception and sensitivity to bodily sensations has been observed in EDs (Sim and Peterson 2021), internalizing disorders (Domschke et al. 2010; Paulus and Stein 2010), and gastrointestinal diseases including both IBD (Vinni et al. 2023) and “functional” disorders such as IBS (Icenhour et al. 2017). There is currently little research on genetic influences on well‐operationalized measures of interoception, though genetic influences have been found on related constructs (e.g., physical anxiety sensitivity; Brown et al. 2012). Additionally, research is needed to understand the ~28% of variance in EDs attributable to unique genetic influences, which may impact molecular/neural mechanisms that specifically contribute to ED risk (e.g., body image perception, appetitive cues).

This study had several strengths, including a large, population‐based sample of women and men, transdiagnostic analyses, and ability to examine genetic/environmental effects using twin models. Nevertheless, some limitations should be noted. We relied on self‐reported diagnoses, and replication is needed using clinician‐assessed diagnoses or other objective measures (e.g., medical records). We also lacked data on current (rather than lifetime) diagnoses. However, phenotypic associations were similar when using a dimensional measure of current ED symptoms (see Tables S6–S7 and Figure S1), suggesting consistency across measurement approaches. There were limitations to some of our covariate measures (e.g., height/weight were self‐reported, income was not weighted to number of household members) and analyses omitted some EDs (e.g., avoidant‐restrictive food intake disorder [ARFID]) because they were not assessed on the MTP questionnaire. It is important for future studies to address these measurement issues and include ARFID, which is associated with increased risk of both gastrointestinal disease (Fink et al. 2022) and internalizing disorders (particularly anxiety disorders; Bourne et al. 2020). We also had lower power to examine specific diagnoses within each diagnostic category, leading to wider confidence intervals for estimates of effects, and were not able to examine OCD, PTSD, or individual ED diagnoses in twin models given smaller sample sizes. Relatively few men had EDs, reducing power to detect sex differences (though notably, the lack of significant sex differences was replicated in analyzes of dimensional ED symptoms, which were better powered to detect such effects). We were not able to estimate qualitative sex differences in etiologic effects due to the small number of opposite sex twin pairs.

While representative of Michigan, our sample had somewhat limited racial and ethnic diversity, and additional research is needed to confirm that associations are similar in people from different backgrounds. It would also be informative to examine associations in youth, as patterns of associations and etiologic mechanisms may differ pre‐puberty. Our data were cross‐sectional, and additional research is needed to assess the typical order of onset across conditions. Further work is needed to verify that associations are not impacted by potential biases in reporting or measurement (e.g., individuals with one known disorder being more likely to attend medical appointments where another may be diagnosed, individuals who are willing to disclose an ED on a survey being more willing to report other conditions). Future research using molecular genetic designs is also important to identify the specific genes and molecular pathways implicated in comorbidity.

Despite these limitations, our findings have important implications for treatment and research. It is imperative to screen individuals with EDs for gastrointestinal disease in addition to comorbid internalizing disorders. Clinicians should conduct a comprehensive mental and physical health evaluation to gain a broader picture of patients' health given significant interplay between EDs and other psychiatric and physical disorders. A more holistic view gained from increased screening could result in more personalized medicine and a more favorable prognosis (Attoe et al. 2018). Eventual identification of shared molecular mechanisms underlying EDs, gastrointestinal disease, and internalizing disorders also has the potential to lead to interventions that alleviate symptoms transdiagnostically.

Author Contributions

Megan E. Mikhail: conceptualization, formal analysis, investigation, methodology, visualization, writing – original draft. Laura A. Pascoe: conceptualization, investigation, methodology, writing – original draft. S. Alexandra Burt: conceptualization, funding acquisition, investigation, methodology, writing – review and editing. Kristen M. Culbert: conceptualization, investigation, methodology, writing – review and editing. Kelly L. Klump: conceptualization, funding acquisition, investigation, methodology, supervision, writing – review and editing.

Ethics Statement

Study procedures were approved by the Michigan State University Institutional Review Board (protocol LEGACY06‐829 M).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.