Abstract
Backgrounds and Aims
Celiac disease (CeD) presents with various gastrointestinal and systemic comorbidities, causing significant social and health burdens. Given the substantial symptom burden and lack of therapeutic options beyond the gluten-free diet, patients with CeD may have limited or unsatisfying interactions with healthcare providers.
Methods
Using the All of Us Research Program, we analyzed five categories of surveys: basic, lifestyle, overall health, social factors of health, and healthcare access and utilization, for participants with self-reported or EHRs-documented CeD. To increase diagnostic accuracy, we excluded CeD patients without compatible HLA haplotypes or with a polygenic risk score (PRS) associated with a low probability of CeD. Each patient with CeD was matched to cohort participants without CeD using the following matching parameters: age, genetically inferred ancestry, and sex.
Results
Individuals with CeD (n = 1816) were more likely to report severe fatigue (12.0% vs 8.4%), frequent specialist visits (≥13 per year: 6.6% vs. 4.3%), and negative interactions with healthcare. On multivariate logistic regression, the following symptoms and healthcare interactions were independently associated with celiac disease: fatigue severity (OR = 1.19, 95% CI [1.11–1.27]), more frequent medical specialist visits OR = 1.69, 95% CI: 1.51–1.90, and negative provider interactions were all positively associated with increased likelihood of CeD.
Conclusions
Participants with celiac disease reported significantly greater functional limitations, higher difficulty assessing care, and a higher degree of fatigue, compared to participants without CeD, underscoring the need to integrate social determinants of health and patient-reported experiences into clinical management.
Keywords: Celiac Disease, Disability, Fatigue, Patient-reported outcomes, Disparities
Introduction
Celiac disease (CeD) is an autoimmune condition affecting the small intestine, triggered by the ingestion of gluten, a protein complex found within commonly consumed cereal grains (i.e., wheat, barley, and rye). CeD affects approximately 1% of the population, with women and non-Hispanic whites being more susceptible to CeD.1,2
The primary treatment for CeD is lifelong adherence to a gluten-free diet, which involves eliminating gluten-containing grains such as wheat, barley, and rye. This often requires avoiding staple foods and maintaining strict vigilance to prevent accidental exposure to gluten. This often results in social consequences for CeD patients, with CeD patients facing social stigma for these precautions and social isolation from being unable to engage in social activities meaningfully.3 Likewise, the monetary cost of maintaining a gluten-free diet is more taxing than a traditional diet, with gluten-free alternatives often costing more.4 Additionally, gluten-free diets often lack certain micronutrients, including Vitamin D, Vitamin B12, folate, and macronutrients such as fiber, which may then require additional costly supplementation.5 While a gluten-free diet is essential for managing the onset of symptoms, it introduces various challenges that affect an individual’s quality of life.
Though physical and psychiatric impairment have been described in prior studies,6,7 data from the United States have been largely limited to single-center analyses at referral centers8 or by claims-based data with limited patient-reported outcomes.9 There remains a need for large-scale and nationally representative studies, including diverse demographics and self-reported data. We, therefore, aimed to assess physical disability and health care interactions using patient-reported data collected in a national cohort study. Our analysis focused on fatigue, functional disability, and healthcare engagement among individuals with and without CeD to identify disparities and inform future care strategies.
Methods
The All of Us Research Program is a U.S. government-funded initiative led by the National Institutes of Health (NIH), designed to recruit from historically underrepresented groups. It is a large cohort study aiming to enroll one million U.S. residents from diverse genetic and cultural backgrounds, with a focus on underrepresented populations. Currently, more than 600,000 participants are enrolled, with detailed data on medical histories, measurements such as height and weight, socioeconomic status, health access, and genetic information. Compared to the UK Biobank, a dataset that includes participants with mostly similar backgrounds and a lack of real-time adaptability,10 the All of Us program gathers data from diverse demographics, including information about patient-reported outcomes and social determinants of health, with ongoing updates with continuous data collection, offering a window into the lives of patients with chronic conditions such as CeD.
For our analysis, we utilized the Version 8 Curated Data Repository (Controlled Tier C2024Q3R4) from the All of Us Research Program, which includes participant data until October 1, 2023. These controlled-tier datasets include data from surveys, electronic health records (EHRs), and genomic data, providing a comprehensive view of participant health and demographics. We identified CeD patients based on either self-reported survey responses (n=1228), a diagnosis documented in the EHR (n=1410), or both (n=842). Genetic validation was established using genetic markers for CeD susceptibility. Specifically, individuals were excluded if they lacked a compatible HLA haplotype (DQ2.5, 2.2, 7.5 or 8.1) or if their polygenic risk score (PRS) for celiac disease was below 3.35. This PRS was calculated using the method by Sharp et al., which incorporates both HLA and non-HLA genetic variants to estimate CeD susceptibility.11 A PRS score of 3.35 corresponds to the best cut-off to discriminate confirmed cases and non-CeD controls using the Youden Index.11 We then matched patients with CeD to up to 5 controls without CeD, utilizing propensity scores based on age, sex, ethnicity, and genetic similarity, using sixteen principal components calculated from single-nucleotide polymorphisms.12
We hypothesized that individuals with CeD would carry a higher symptom burden and healthcare usage than non-CeD individuals, with differences across racial and ethnic groups. Due to the usage of multiple comparisons, secondary analyses were FDR-adjusted rather than Bonferroni-adjusted in order to preserve statistical power and control false discoveries.
We compared CeD patients to matched controls with regard to self-reported answers on the following topics: basic demographics and disability status—The Basics (n = 8), Lifestyle (n = 7), Overall Health (n = 5), Social Factors of Health (n = 12), and Healthcare Access and Utilization (n = 3). At the time of enrollment, participants complete “The Basics” questionnaire, which is required for the participant to be eligible to continue the remaining All of Us protocol. “The Basics” include demographic and identity information, such as age, sex-at-birth, race, ethnicity, education level, income, and disability. Following this questionnaire, participants are asked to complete another set of surveys–the “Overall Health” and “Lifestyle” questionnaires. The “Overall Health” questionnaire captured general well-being and fatigue characteristics (e.g., “In the past 7 days, how would you rate your fatigue?” and “In general, how would you rate your physical health?”). The “Lifestyle” questionnaire included questions about the participants' use of tobacco, alcohol, and other drugs (e.g, “On a typical day when you drink, how many drinks do you have?” and “Have you ever used an electronic nicotine product, even one or two times?”). After 90 days of enrollment, participants are provided with the “Healthcare Access & Utilization” questionnaire.13 These surveys are designed to acquire more detailed and longitudinal data, building on the initial demographic information. The “Healthcare Access & Utilization” module captures the frequency of healthcare interactions and barriers to care. including questions such as “What is the total number of general doctor visits you made in the last 12 months?”. The data in this analysis were gathered by All of Us from participants via a set of questionnaires during enrollment, with participants being able to skip questions and surveys not included in All of Us’s “core surveys” (The Basics, Lifestyle, and Overall Health).13
We compared CeD patients to controls with regard to questionnaire responses using chi-squared tests, performed on All of Us Researcher Workbench. False discovery rate (FDR) correction was applied to account for multiple comparisons (n=4679), and statistical significance was defined as an FDR-corrected p-value < 0.05. In these comparisons, we excluded responses of “Prefer Not To Answer” and “Don’t Know”. The results are reported in accordance with the All of Us Data and Statistics Dissemination Policy and are displayed only for groups of at least 20 individuals. We then constructed a multivariable model to identify responses that were independently associated with CeD. Variables with a significant bivariate association were evaluated after calculating the variance inflation factor (VIF) scores to test for multicollinearity among variables, and mapping ordinal variables (such as those based on rating scales) to numeric scales in order to preserve their order. We used an iterative model refinement approach to identify a parsimonious set of significant predictors. All preprocessing of data and statistical analyses were performed on Python using the Statsmodels and Pandas libraries, and the logistic regression was conducted using the Broyden-Fletcher-Goldfarb-Shanno (BFGS) method, which is commonly used for efficiently estimating model parameters. For the multivariable analysis, we used raw two-sided p-values < 0.05 as the significance cutoff.
Results
Identification of CeD patients and Controls
Among All of Us participants with a diagnosis of CeD (n=3480) recorded in the EHR (n=1410), self-reported (n=1228) and both (n = 842), 581 (18.9%) were excluded because of incompatible HLA haplotypes, and 1083 (35.2%) were excluded because of a low probability of CeD based on a polygenic risk score of < 3.35. This resulted in a final analytical cohort of 1816 CeD patients that were genetically confirmed using both HLA haplotypes and polygenic risk scores, matched to 9080 controls based on age, sex, and genetic ancestry (see Figure 1).
Figure 1:
Matching Procedure Flowchart
Sociodemographic Differences
Notable differences in sociodemographic characteristics were observed between individuals with and without CeD (see Table 1). Among participants with CeD, a larger number of individuals reported being white, with 1654 (91.1%), compared to 7891 (86.9%) among controls. Educational attainment was higher for CeD patients compared to controls (p = 0.0271). Individuals with CeD were more likely to hold an advanced degree (35.0%), compared to controls (27.0%). The proportions of college graduates (28.8% vs 27.8%) and individuals with some college educations (24.1% vs 26.0%) were similar between the groups. In contrast, a smaller percentage of CeD participants reported an educational attainment of a high school diploma or GED (9.4% vs 13.9%).
Table 1:
Demographics and Lifestyle by CeD Status
| Variable | CeD Count (%) | Control Count (%) | P Value |
|---|---|---|---|
|
| |||
| Race/Ethnicity | CeD (n=1816) | Control (n=9080) | 0.02709 |
| White | 1654 (91.1%) | 7891 (86.9%) | |
| Black | 45 (2.5%) | 165 (1.8%) | |
| Hispanic | 125(6.9%) | 850 (9.4%) | |
| AIAN | 57 (3.1%) | 269 (3.0%) | |
| Other | 47 (2.6%) | 328 (3.6%) | |
|
| |||
| Education Level | CeD (n=1816) | Control (n=9080) | <0.0001 |
| High School/GED | 171 (9.4%) | 1260 (13.9%) | |
| Some College | 438 (24.1%) | 2358 (26.0%) | |
| College Graduate | 523 (28.8%) | 2527 (27.8%) | |
| Advanced Degree | 635 (35.0%) | 2454 (27.0%) | |
|
| |||
| Has Health Insurance | CeD (n=1816) | Control (n=9080) | <0.0001 |
| High School/GED | 1776 (97.8%) | 8631 (95.1%) | |
|
| |||
| Income Group | CeD (n=1816) | Control (n=9080) | <0.0001 |
| <10,000 | 85 (4.7%) | 689 (7.6%) | |
| $10k-25k | 146 (8.0%) | 853 (9.4%) | |
| $25k-35k | 109 (6.0%) | 585 (6.4%) | |
| $35k-50k | 175 (13.3%) | 802 (13.2%) | |
| $50k-75k | 241 (14.0%) | 1201 (10.7%) | |
| $75k-100k | 255 (14.4%) | 974 (14.1%) | |
| $100k-150k | 261 (7.2%) | 1279 (6.8%) | |
| $150k+ | 341 (18.8%) | 1397 (15.4%) | |
|
| |||
| Employment Status | CeD (n=1816) | Control (n=9080) | 0.0135 |
| Employment for Wages | 852 (46.9%) | 4270 (47.1%) | |
| Self-Employed | 146 (8.0%) | 730 (8.0%) | |
| Out of Work ≥ 1 yr | 68 (3.7%) | 412 (4.5%) | |
| Out of Work < 1yr | 50 (2.8%) | 301 (3.3%) | |
| Homemaker | 85 (4.7%) | 587 (6.5%) | |
| Student | 125 (6.9%) | 504 (5.6%) | |
| Retired | 501 (27.6%) | 2201 (24.2%) | |
| Unable to Work | 195 (10.7%) | 838 (9.2%) | |
|
| |||
| Marital Status | CeD (n=1816) | Control (n=9080) | 0.0364 |
| Married | 982 (54.1%) | 4553 (50.1%) | |
| Divorced | 226 (12.4%) | 1311 (14.4%) | |
| Widowed | 76 (4.2%) | 484 (5.3%) | |
| Separated | 28 (1.5%) | 223 (2.5%) | |
| Never Married | 355 (19.5%) | 1725 (19.0%) | |
|
| |||
| Insurance Type | CeD (n=1658) | Control (n=8151) | 0.004514 |
| Employer/Union | 908 (54.8%) | 4215 (51.7%) | |
| Medicare | 500 (30.2%) | 2273 (27.9%) | |
| Medicaid | 230 (13.9%) | 1535 (18.8%) | |
| Military | 43 (2.6%) | 232 (2.9%) | |
| Purchased | 182 (11.0%) | 839 (10.3%) | |
| VA | 73 (4.4%) | 286 (3.5%) | |
| Other/None | 52 (3.1%) | 233 (2.9%) | |
While health insurance did not differ considerably between groups, household income distributions varied (p = 0.0001). Participants with CeD were less likely to report incomes in the lowest brackets: 4.7% of the CeD groups reported earning less than the $10,000 annual range compared to 7.6% of controls, and 8.0% fell within the $10,000-$25,000 range versus 9.4% of controls. On the other end of the income spectrum, 18.8% reported incomes of $150,000 and above compared to 15.4% of controls. Moreover, there was a modest gap in employment status. Of the 1816 participants in the CeD group, 696 (38.3%) reported being retired or unable to work, compared to 3039 (33.5%) of controls (p < 0.0001).
General Health and Disability Burden
CeD participants reported a higher burden of physical limitations and fatigue. Among 1423 CeD participants, limitations in activities of daily living—specifically, difficulty dressing/bathing—appeared to be more common in individuals with CeD (5.2%) than in non-CeD controls (3.3%, p = 0.0324, see Table 2).
Table 2:
Overall Health and Disability by CeD Status
| Variable | CeD Count (%) | Control Count (%) | P Value |
|---|---|---|---|
|
| |||
| Activities of Daily Living | CeD (n=1423) | Control (n=6696) | 0.0324 |
| Unable to Dressing/Bathing | 74 (5.2%) | 223 (3.3%) | |
|
| |||
| Average Fatigue: Past 7 Days | CeD (n=1815) | Control (n=8955) | <0.0001 |
| None | 257 (14.2%) | 1571 (17.5%) | |
| Mild | 675 (37.2%) | 3800 (42.3%) | |
| Moderate | 606 (33.4%) | 2626 (29.2% | |
| Severe | 218 (12.0%) | 757 (8.4%) | |
| Very Severe | 52 (2.9%) | 163 (1.8%) | |
|
| |||
| General Health | CeD (n=1815) | Control (n=8991) | <0.0001 |
| Excellent | 142 (7.8%) | 1001 (11.1%) | |
| Very Good | 606 (33.4%) | 3107 (34.6%) | |
| Good | 599 (33.0%) | 3047 (33.9%) | |
| Fair | 349 (19.2%) | 1425 (15.8%) | |
| Poor | 108 (6.0%) | 346 (3.9%) | |
|
| |||
| General Physical Health | CeD (n=1815) | Control (n=8991) | 0.0018 |
| Excellent | 128 (7.1%) | 852 (9.5%) | |
| Very Good | 556 (30.6%) | 2817 (31.3%) | |
| Good | 618 (34.1%) | 3142 (35.0%) | |
| Fair | 384 (21.2%) | 1649 (18.3%) | |
| Poor | 119 (6.6%) | 531 (5.9%) | |
|
| |||
| Everyday Activity | CeD (n=1815) | Control (n=8991 | 0.0072 |
| Not at all/A Little | 162 (8.9%) | 629 (7.0%) | |
| Moderately | 216 (11.9%) | 977 (10.9%) | |
| Mostly | 311 (17.1%) | 1468 (16.5%) | |
| Completely | 1116 (61.5%) | 5848 (65.0%) | |
Fatigue severity was greater among individuals with CeD compared to controls (p < 0.0001). A lower proportion of patients with CeD inhabited the no fatigue category — 14.2% compared to 17.5%. For the mild fatigue category, 37.2% of the CeD group reported mild fatigue compared to 42.3% of the control group. Instead, CeD patients reported higher representation in more severe categories, with 12.0% of the CeD group experiencing severe fatigue versus 8.4% of controls.
Perceptions of general health were also less favorable among CeD participants. CeD participants were significantly less likely to rate their health as “excellent” and more likely to report “fair” or “poor” health (p < 0.0001, see Table 2). This also applied to general physical health, where only 128 (7.05%) of CeD individuals rated their status as “excellent” compared to 852 (9.48%) of controls, while “fair” or “poor” ratings were again more frequently reported in the CeD group (p = 0.002).
Daily functioning differed significantly among individuals with CeD and controls, with the distribution of functional capacity ratings differing meaningfully between groups. In contrast, a majority in both groups reported being able to carry out everyday activities “completely,” the proportion was lower among CeD participants (61.5%) than controls (65.0%). 61.5% of participants with CeD reported having a completely daily functional capacity compared to 65% in controls (p = 0.007).
Healthcare Access and Patient-Provider Experiences
Healthcare utilization patterns varied significantly between individuals with CeD and non-CeD controls across various provider categories. CeD participants reported more frequent visits to physicians, with a notably smaller proportion falling in the lowest visit bracket (52.1%) compared to controls (59.1%, p < 0.0001, see Table 3). Individuals with CeD reported significantly higher rates of interaction with physicians and medical specialists, compared to controls.
Table 3:
Interactions with Health Care by CeD Status
| Variable | CeD Count (%) | Control Count (%) | P Value |
|---|---|---|---|
|
| |||
| Doctor/Nurse Doesn’t Listen | CeD (n=1201) | Control (n=5012) | 0.000532 |
| Never | 361 (30.1%) | 1757 (35.1%) | |
| Rarely | 333 (27.7%) | 1514 (30.2%) | |
| Sometimes | 367 (30.6%) | 1301 (26.0%) | |
| Most of the time or Always | 118 (9.8%) | 354 (7.6%) | |
|
| |||
| Doctor/Nurse Acts Better Than You | CeD (n=1201) | Control (n=5010) | 0.00552 |
| Never | 583 (48.5%) | 2741 (54.7%) | |
| Rarely | 329 (27.4%) | 1285 (25.6%) | |
| Sometimes | 206 (17.2%) | 728 (14.5%) | |
| Most of the time or Always | 64 (5.3%) | 175 (3.5%) | |
|
| |||
| Doctor/Nurse Thinks You’re Not Smart | CeD (n=1200) | Control (n=5011) | 0.0062 |
| Never | 647 (53.9%) | 2944 (58.8%) | |
| Rarely | 291 (24.3%) | 1186 (23.7%) | |
| Sometimes | 192 (16.0%) | 579 (11.6%) | |
| Most of the time or Always | 48 (4.0%) | 166 (3.3%) | |
|
| |||
| Number of Doctor Visits | CeD (n=1297) | Control (n=5072) | <0.0001 |
| 1–3 | 647 (52.1%) | 2996 (59.1%) | |
| 4–7 | 342 (26.4%) | 1291 (25.5%) | |
| 8–12 | 108 (8.3%) | 365 (7.2%) | |
| 13–15 | 27 (2.1%) | 78 (1.5%) | |
| 16+ | 118 (9.1%) | 256 (5.0%) | |
|
| |||
| Number of Medical Specialist Visits | CeD (n=944) | Control (n=2925) | 0.001649 |
| 1–3 | 520 (55.1%) | 1834 (62.7%) | |
| 4–7 | 266 (28.2%) | 716 (24.5%) | |
| 8–12 | 82 (8.7%) | 199 (6.8%) | |
| 13+ | 62 (6.6%) | 125 (4.3%) | |
|
| |||
| Respected By Provider | CeD (n=1434) | Control (n=5788) | 0.0034 |
| Always | 939 (65.5%) | 4079 (70.5%) | |
| Usually | 410 (28.6%) | 1448 (25.0%) | |
| Some of the time/None | 79 (5.5%) | 212 (3.7%) | |
|
| |||
| Spoken to Professional | CeD (n=1463) | Control (n=6091) | <0.0001 |
| 6 Months or Less | 1296 (88.6%) | 5079 (83.4%) | |
| 6mo-1yr | 121 (8.3%) | 656 (10.8%) | |
| More Than One Year Ago | 29 (2.0%) | 292 (4.8%) | |
With regard to healthcare interactions, CeD patients reported feeling less respected by their healthcare providers than controls (Table 3). These included responses to the following questions: “How often does your [healthcare provider] act as if she is better than you?”, “How often do you feel like your [healthcare provider] is not listening to what you are saying?”, and “How often does a [healthcare provider] act if he or she thinks you are not smart…?”. To all of these questions, CeD patients responded “never” less frequently than controls (48.5% vs 54.7%, 30.1% vs. 35.1%, and 53.9% vs 58.8% respectively, see Table 3).
Multivariable Analysis
The final multivariable logistic regression model (Table 4) included 10,894 observations and provided a poor fit to the data (McFadden’s R2 = 0.05). Of the 34 variables initially input into the regression, 15 remained significant.
Table 4:
Logistic Regression Results
| Variable | Odds Ratio | 95% Confidence Interval | P Value |
|---|---|---|---|
|
| |||
| Ordinal | |||
| Overall Health: Average Fatigue | 1.19 | 1.11–1.27 | <0.0001 |
| Education Level: Highest Grade | 1.14 | 1.08–1.20 | <0.0001 |
| Overall Health: General Health | 1.10 | 1.04–1.18 | 0.002 |
| Doctor/Nurse Thinks You’re Not Smart | 1.10 | 1.03–1.17 | 0.006 |
| Health Advise: General Doctor Visits | 1.05 | 1.02–1.09 | 0.003 |
| Alcohol: Average Daily Drink Count | 0.79 | 0.70–0.89 | 0.0001 |
| Smoking: Smoke Frequency | 0.75 | 0.66–0.85 | <0.0001 |
|
| |||
| Nominal | |||
|
| |||
| Spoken to Medical Specialist | |||
| Yes | 1.69 | 1.51–1.90 | <0.0001 |
| No | 1.00 | [reference] | - |
|
| |||
| Spoken to Podiatrist | |||
| Yes | 1.30 | 1.09–1.54 | 0.004 |
| No | 1.00 | [reference] | - |
|
| |||
| Place for Health Advice | |||
| Yes | 1.19 | 1.02–1.38 | 0.023 |
| No | 1.00 | [reference] | - |
|
| |||
| Health Insurance | |||
| Yes | 1.00 | [reference] | - |
| No | 0.62 | 0.41–0.94 | 0.026 |
|
| |||
| Insurance Accepted | |||
| Yes | 1.19 | 1.01–1.40 | 0.041 |
| No | 1.00 | [reference] | - |
|
| |||
| Insurance Type Update | |||
| Employer or Union | 1.00 | [reference] | - |
| Medicaid | 0.81 | 0.68–0.96 | 0.017 |
|
| |||
| Employment Status | |||
| Student | 1.27 | 1.03–1.57 | 0.028 |
| Retired | 1.16 | 1.03–1.32 | 0.017 |
| Employed For Wages | 1.00 | [reference] | - |
| Homemaker | 0.75 | 0.59–0.95 | 0.019 |
|
| |||
| Martial Status | |||
| Married | 1.00 | [reference] | - |
| Divorced | 0.84 | 0.72–0.99 | 0.034 |
| Widowed | 0.70 | 0.54–0.91 | 0.007 |
Among the ordinal variables, the strongest association was for fatigue (OR = 1.19, 95% CI: 1.11–1.27, p < 0.0001). Higher education level was also positively associated with CeD (OR = 1.14, 95% CI: 1.08–1.20, p < 0.0001). In contrast, higher smoking frequency and daily alcohol consumption were associated with decreased odds of CeD (Smoking: OR = 0.75, 95% CI: 0.66–0.85, p < 0.0001; Alcohol: OR = 0.79, 95% CI: 0.70–0.89, p = 0.0001). Seeking specialist care and having multiple sources for health advice were associated with CeD.
Regarding insurance-related variables, lack of health insurance was associated with a 38% reduction in odds compared to those with insurance (OR = 0.62, 95% CI: 0.41–0.94, p = 0.026). Among insured participants, having insurance that was accepted increased the odds of CeD compared to those whose insurance was denied (OR = 1.19, 95% CI: 1.01–1.40, p = 0.041). Individuals covered by Medicaid had significantly lower odds of CeD compared to those with employer- or union-based insurance (OR = 0.81, 95% CI: 0.68–0.96, p = 0.017).
The association between CeD and having the frequent feeling that “a [healthcare provider] acts if he or she thinks you are not smart” remained significant on multivariable analysis (OR 1.10; 95% CI 1.03–1.10, p = 0.006).
Discussion
In this study, we assessed the burden of CeD beyond symptoms and the gluten-free diet and found significant disparities in disability, general health perception, healthcare visits, and negative impressions of healthcare interactions. As these aspects of CeD are often overlooked, these findings underscore the value of developing a more holistic framework for the management of CeD.
Very little is known about the socioeconomic makeup of CeD patients in the United States. Patients with a lower socioeconomic status are more likely to have diarrhea as their presenting symptom, possibly because those with non-classical presentations may remain undiagnosed.14 In the present analysis, CeD participants were more likely to hold an advanced degree compared to non-CeD controls. CeD participants were disproportionately represented in the middle- and upper-income brackets. This may reflect both disparities in healthcare access and diagnostic biases; populations from marginalized racial and socioeconomic demographics may be affected by structural barriers, which can be related to limited access to diagnostic tests, medical specialists, or decreased health awareness about gluten-related conditions. At the same time, CeD patients reported lower rates of participation in the workforce, suggesting that the condition may be associated with economic burdens. There were also higher rates for CeD participants in the “Retired” (27.6% vs 24.2%) and “Unable to Work” (10.7% vs 9.2%) categories, which may reflect the greater experience of average fatigue and psychological distress, along with systemic symptoms that affect sustained employment. Additionally, disability can be associated with an inability to work, as functional impairments can limit an individual’s ability to perform occupational tasks or maintain employment. National data show that only 38.4% of working-age adults (18–64 years old) with disabilities are employed compared to 75.8% who do not have disabilities.15
These patterns suggest that while CeD is more often diagnosed in populations with access to resources, it may still impose a substantial functional toll on those who are affected.
We found that fatigue is a substantial problem in patients with CeD (see Figure 2). The etiology of fatigue is likely multifactorial and may include chronic inflammation and nutritional deficiencies from small intestinal villus atrophy. However, participants with CeD have experienced severe fatigue even after adoption of the gluten-free diet,16 which suggests that reversal of inflammation and villus atrophy is insufficient to address this common symptom. Fatigue is frequent and closely linked to depression in CeD, which improved with a gluten-free diet; it may persist and not be independently associated with gastrointestinal symptoms or nutritional status.20, 21 Fatigue is a particularly common and impactful extraintestinal symptom, which is present at diagnosis and associated with declined quality of life.18, 19 These results support the integration of functional health assessments and fatigue screenings into routine CeD management to capture the disease burden and track the progress of therapeutics.
Figure 2:
Distribution of fatigue severity among participants with and without Celiac Disease (CeD). This bar graph shows the percentage of participants who experience each level of fatigue: None, Mild, Moderate, Severe, and Very Severe. CeD participants reported higher levels of severe and very severe fatigue, while Non-CeD participants were more likely to report “None” and “Mild”. These findings highlight an increase in fatigue burden among individuals with celiac disease.
We also explored the broader impact of CeD on perceived general health and patterns of healthcare utilization. Our data revealed a consistent trend among CeD participants of reduced self-perceived health. The CeD group was less likely to rate their general or physical health as “excellent” and more likely to report “fair” or “poor” health. This discrepancy may reflect the logistical and emotional burdens that CeD places on individuals. Perhaps the most notable results in this analysis pertain to the perception among CeD patients of being dismissed by their healthcare providers. CeD responses regarding healthcare interactions included the impression that their providers did not listen to them, acted better than them, or did not think they were smart [Figure 3]. Previous studies have reported that patients often perceive limited provider knowledge regarding CeD and inadequate time spent addressing their concerns.22 However, those investigations employed the Coeliac Disease Assessment Questionnaire (CDAQ), while our analysis utilized data from All of Us Questions and Answer surveys. Despite having higher levels of educational attainment compared to non-CeD controls, participants with CeD reveal feeling more dismissed and undermined in both their treatment and in their understanding; indeed, the impression that their healthcare provider feels they are “not smart” was reported more commonly in CeD patients, a finding that remained significant on multivariable analysis. These experiences were notable given their frequency of engagement with healthcare providers, their higher rates of visits with specialists, and their higher level of educational attainment. This may suggest a failure in trust and communication and/or the necessity for provider awareness of the challenges facing patients with CeD.
Figure 3:
CeD participants consistently reported feeling less respected during their interactions with providers. These bar graphs visualize the mean (with SD) participants who responded with “Never” and “Most of the Time or Always” on questions which asked “How often does your [healthcare provider] act as if she is better than you”, “How often do you feel like your [healthcare provider] is not listening to what you are saying”, and “How often does a [healthcare provider] act if he or she thinks you are not smart”.
Strengths of this study include its broad geographic and demographic reach in the United States, as the All of Us dataset intentionally includes populations that are underrepresented racially, ethnically, and socioeconomically. This, therefore, allows for the investigation of CeD to reach a broader demographic and help reveal disparities in diagnosis, experience, and self-perception. This study also has some limitations. The diagnosis of CeD was identified by self-reported survey responses and diagnoses in the EHR, and confirmation via serology and intestinal biopsy was not available. (Though elevated tissue transglutaminase IgA was an identifiable field in the electronic health record, this is a marker in newly diagnosed CeD, and is negative in most prevalent cases.) To mitigate the possibility of false-positive cases, we restricted the CeD population to those with a permissive HLA haplotype and a sufficiently high polygenic risk score. Data on adherence to the gluten-free diet were not available, nor was there information on mucosal healing, which may affect the risk of long-term morbidity in CeD.17 Additionally, this analysis did not account for potential comorbid conditions—such as psychiatric disorders—that may influence key outcomes like daily functioning and health-related quality of life. Since fatigue is multifactorial, there can be potential contributors, such as sleep disturbance, which were not consistently present in the dataset and could not be included. These contributors can affect fatigue severity and should be considered when interpreting these results. Future research should investigate how these factors can be used to predict long-term adherence, healthcare utilization, and mental health outcomes. There is also a potential for selection bias in identifying patients with CeD based in clinical testing as individuals who undergo serological testing and/or biopsy may be different from individuals who remain undiagnosed in the long-term. CeD status in All of Us is based on documented ICD-10 codes, which raises the possibility that individuals represented as controls may have undiagnosed CeD. This risk of misclassification can be due to structural barriers to diagnosis which is common in racially and ethnically minoritized groups and should be considered when interpreting comparisons between groups. Furthermore, integrating additional qualitative methods into future studies, such as interviews with CeD patients across clinical and community settings, could deepen insights into the experiences of living with this condition.
Conclusion
This analysis of a large cohort found that CeD has a substantial burden on individuals, affecting fatigue, daily functioning, and healthcare interactions in the U.S. Our findings highlight the importance of a multidisciplinary approach to CeD care, which should encompass both psychological and patient communication models. Future longitudinal studies can investigate how these outcomes change over time and how tailored interventions can alleviate the burden of CeD. As CeD places a significant burden on affected individuals in the domains of fatigue, disability, and healthcare interactions, future efforts should focus on the development of personalized and equitable care strategies for individuals with CeD.
Acknowledgments
The All of Us research program is supported by the National Institutes of Health, Office of the Director: Regional Medical Centers: 1 OT2 OD026549; 1 OT2 OD026554; 1 OT2 OD026557; 1OT2 OD026556; 1 OT2 OD026550; 1 OT2 OD 026552; 1 OT2 OD026553; 1 OT2 OD026548; 1OT2 OD026551; 1 OT2 OD026555; IAA #: AOD 16037; Federally Qualified Health Centers: HHSN 263201600085U; Data and Research Center: 5 U2C OD023196; Biobank: 1 U24OD023121; The Participant Center: U24 OD023176; Participant Technology Systems Center: 1U24 OD023163; Communications and Engagement: 3 OT2 OD023205; 3 OT2 OD023206; and Community Partners: 1 OT2 OD025277; 3 OT2 OD025315; 1 OT2 OD025337; 1 OT2OD025276. The All of Us Research Program would not be possible without the partnership established with its participants.
Funding
Dr. Xiao-Fei Kong is supported by the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health under Award Number K08DK128631 and Disease-Oriented Clinical Scholars Program at the UT Southwestern Medical Center.
Footnotes
Disclosures: The authors have no relevant financial or non-financial interests to disclose.
Statement: The data and code used in this study are available as a shared workspace to registered researchers of the All of Us Researcher Workbench. For information about access, please visit https://www.researchallofus.org/.
References
- [1].Choung RS, Ditah IC, Nadeau AM, Rubio-Tapia A, Marietta EV, Brantner TL, Camilleri MJ, Rajkumar SV, Landgren O, Everhart JE, Murray JA. Trends and racial/ethnic disparities in gluten-sensitive problems in the United States: findings from the National Health and Nutrition Examination Surveys from 1988 to 2012. Am J Gastroenterol. 2015. Mar;110(3):455–61. doi: 10.1038/ajg.2015.8. Epub 2015 Feb 10. PMID: 25665935. [DOI] [PubMed] [Google Scholar]
- [2].Mardini HE, Westgate P, & Grigorian AY (2015). Racial Differences in the Prevalence of Celiac Disease in the US Population: National Health and Nutrition Examination Survey (NHANES) 2009–2012. Digestive Diseases and Sciences, 60(6). 10.1007/s10620-014-3514-7 [DOI] [PubMed] [Google Scholar]
- [3].Lebovits J, Lee AR, Ciaccio EJ, Wolf RL, Davies RH, Cerino C, Lebwohl B, Green PHR. Impact of Celiac Disease on Dating. Dig Dis Sci. 2022. Nov;67(11):5158–5167. doi: 10.1007/s10620-022-07548-y. Epub 2022 May 30. PMID: 35635630; PMCID: PMC9150385. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Lee AR, Wolf RL, Lebwohl B, Ciaccio EJ, Green PHR. Persistent Economic Burden of the Gluten Free Diet. Nutrients. 2019. Feb 14;11(2):399. doi: 10.3390/nu11020399. PMID: 30769836; PMCID: PMC6412592. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Lee AR, Dennis M, Lebovits J, Welstead L, Verma R, Therrien A, Lebwohl B. Dietary assessments in individuals living with coeliac disease: key considerations. J Hum Nutr Diet. 2025. Feb;38(1):e13380. doi: 10.1111/jhn.13380. Epub 2024 Nov 5. PMID: 39501424; PMCID: PMC11589401. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Skjellerudsveen BM, Omdal R, & Grimstad T (2019). Fatigue in celiac disease: A review of the literature. In JGH Open (Vol. 3, Issue 3). 10.1002/jgh3.12134 [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Lebwohl B, Haggård L, Emilsson L, Söderling J, Roelstraete B, Butwicka A, Green PHR, & Ludvigsson JF (2021). Psychiatric Disorders in Patients With a Diagnosis of Celiac Disease During Childhood From 1973 to 2016. Clinical Gastroenterology and Hepatology, 19(10). 10.1016/j.cgh.2020.08.018 [DOI] [PubMed] [Google Scholar]
- [8].Roldan GA, Jamot S, Kopec K, Charoen A, Leffler D, Feller ER, & Shah SA (2023). Celiac Disease Presenting in a Community-Based Gastroenterology Practice: Obesity and Bone Disease Are Common. Digestive Diseases and Sciences, 68(3). 10.1007/s10620-022-07521-9 [DOI] [PubMed] [Google Scholar]
- [9].Cappell K, Taylor A, Johnson BH, Gelwicks S, Wang S, Gerber M, Leffler DA. Healthcare Resource Utilization and Costs in Celiac Disease: A US Claims Analysis. Am J Gastroenterol. 2020. Nov;115(11):1821–1829. doi: 10.14309/ajg.0000000000000759. PMID: 33156101; PMCID: PMC7725140. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, Downey P, Elliott P, Green J, Landray M, Liu B, Matthews P, Ong G, Pell J, Silman A, Young A, Sprosen T, Peakman T, & Collins R (2015). UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age. PLoS Medicine, 12(3). 10.1371/journal.pmed.1001779 [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Sharp SA, Jones SE, Kimmitt RA, Weedon MN, Halpin AM, Wood AR, Beaumont RN, King S, van Heel DA, Campbell PM, Hagopian WA, Turner JM, & Oram RA (2020). A single nucleotide polymorphism genetic risk score to aid diagnosis of coeliac disease: a pilot study in clinical care. Alimentary Pharmacology and Therapeutics, 52(7). 10.1111/apt.15826 [DOI] [PubMed] [Google Scholar]
- [12].Karnati Hemanth, et al. Integrative Modeling of Genetic and Clinical Risks to Improve Celiac Disease Prediction Across Diverse Populations. Available at SSRN: https://ssrn.com/abstract=5243512 or 10.2139/ssrn.5243512 [DOI]
- [13].Ramirez AH, Sulieman L, Schlueter DJ, Halvorson A, Qian J, Ratsimbazafy F, Loperena R, Mayo K, Basford M, Deflaux N, Muthuraman KN, Natarajan K, Kho A, Xu H, Wilkins C, Anton-Culver H, Boerwinkle E, Cicek M, Clark CR, … Zwick ME (2022). The All of Us Research Program: Data quality, utility, and diversity. Patterns, 3(8). 10.1016/j.patter.2022.100570 [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Roy A, Mehra S, Kelly CP, Tariq S, Pallav K, Dennis M, Peer A, Lebwohl B, Green PH, Leffler DA. The association between socioeconomic status and the symptoms at diagnosis of celiac disease: a retrospective cohort study. Therap Adv Gastroenterol. 2016. Jul;9(4):495–502. doi: 10.1177/1756283X16637532. Epub 2016 Mar 18. PMID: 27366218; PMCID: PMC4913330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Shockey TM, Fox K, Zhao G, Hollis N. Prevalence of Disability by Occupation Group - United States, 2016–2020. MMWR Morb Mortal Wkly Rep. 2023. May 19;72(20):540–546. doi: 10.15585/mmwr.mm7220a1. PMID: 37200224; PMCID: PMC10205170. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Skjellerudsveen BM, Omdal R, Hetta AK, Kvaløy JT, Aabakken L, Skoie IM, Grimstad T. Less, but not gone-gluten-free diet effects on fatigue in celiac disease: a prospective controlled study. Front Med (Lausanne). 2023. Sep 12;10:1242512. doi: 10.3389/fmed.2023.1242512. PMID: 37766922; PMCID: PMC10520960. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Ludvigsson JF, Yao J, Lebwohl B, Green PHR, Yuan S, Leffler DA. Coeliac disease: complications and comorbidities. Nat Rev Gastroenterol Hepatol. 2025. Apr;22(4):252–264. doi: 10.1038/s41575-024-01032-w. Epub 2025 Jan 28. PMID: 39875649. [DOI] [PubMed] [Google Scholar]
- [18].Wieser H, Ciacci C, Soldaini C, Gizzi C, Santonicola A. Gastrointestinal and Hepatobiliary Manifestations Associated with Untreated Celiac Disease in Adults and Children: A Narrative Overview. J Clin Med. 2024. Aug 5;13(15):4579. doi: 10.3390/jcm13154579. PMID: 39124845; PMCID: PMC11312480. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Santonicola A, Wieser H, Gizzi C, Soldaini C, Ciacci C. Associations between Celiac Disease, Extra-Gastrointestinal Manifestations, and Gluten-Free Diet: A Narrative Overview. Nutrients. 2024. Jun 9;16(12):1814. doi: 10.3390/nu16121814. PMID: 38931169; PMCID: PMC11206979. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Jelsness-Jørgensen LP, Bernklev T, Lundin KEA. Fatigue as an Extra-Intestinal Manifestation of Celiac Disease: A Systematic Review. Nutrients. 2018. Nov 3;10(11):1652. doi: 10.3390/nu10111652. PMID: 30400298; PMCID: PMC6266448. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Siniscalchi M, Iovino P, Tortora R, Forestiero S, Somma A, Capuano L, Franzese MD, Sabbatini F, Ciacci C. Fatigue in adult coeliac disease. Aliment Pharmacol Ther. 2005. Sep 1;22(5):489–94. doi: 10.1111/j.1365-2036.2005.02619.x. PMID: 16128688. [DOI] [PubMed] [Google Scholar]
- [22].Crocker H, Jenkinson C, Peters M. Healthcare experiences and quality of life of adults with coeliac disease: a cross-sectional study. J Hum Nutr Diet. 2020. Dec;33(6):741–751. doi: 10.1111/jhn.12757. Epub 2020 May 7. PMID: 32383338. [DOI] [PubMed] [Google Scholar]