This study used the UK Biobank cohort, a population-based large prospective cohort including over 500,000 participants (aged 38–73 years) recruited between 2006 and 2010 from 22 assessment centers across the United Kingdom. Participants completed a self-administered touchscreen questionnaire on sociodemographic characteristics, lifestyle exposures, medical history, medication use, and physical measurements at recruitment. Extensive medical information was collected during both recruitment and the longitudinal follow-up period. Additional details about this cohort can be found elsewhere or on the official website (www.ukbiobank.ac.uk). To analyze the association of gastrointestinal diseases with incident endocrine diseases, we included 481841 participants after excluding participants who had prevalent endocrine diseases at baseline. Similarly, 452858 participants were available when analyzing the association between endocrine diseases and risk of gastrointestinal diseases, and participants who had prevalent gastrointestinal diseases at baseline were excluded. All participants included provided signed informed consent.

The diseases were ascertained based on diagnoses derived from the self-reported questionnaire, and linked data from primary care and hospital inpatient registers, using the International Classification of Diseases tenth edition (ICD-10). For gastrointestinal diseases, besides considering all of them as one outcome, we additionally identified 9 individual diagnoses of gastrointestinal non-neoplastic diseases (gastric ulcer, duodenal ulcer, gastritis and duodenitis, dyspepsia, Crohn’s disease, ulcerative colitis, irritable bowel syndrome, intestinal malabsorption, and gastrointestinal hemorrhage) and 3 individual diagnoses of gastrointestinal neoplastic diseases (gastric cancer, small intestine cancer, and colorectal cancer). For, endocrine diseases, besides considering all of them as one outcome, we additionally identified 12 individual diagnoses of endocrine diseases (hypothyroidism, hyperthyroidism, thyroiditis, T2D, hypoparathyroidism, hyperparathyroidism, hyperpituitarism, hypopituitarism, Cushing syndrome, hyperaldosteronism, ovarian dysfunction, and testicular dysfunction), and 9 subtypes of endocrine diseases according to the involved gland (thyroid, pancreatic, parathyroid, pituitary, adrenal, genital glands) or axis (hypothalamic-pituitary-adrenal [HPA], hypothalamic-pituitary-thyroid [HPT] and hypothalamic-pituitary-gonadal [HPG] axis). A detailed description of the disease definition is provided in Supplementary Tables S1 , S2 .

The covariates used in this study included baseline demographics, lifestyle behaviors, and social isolation. Baseline demographics included age, sex, body mass index (BMI), waist-to-hip ratio (WHR), the Townsend Deprivation Index (a composite index reflecting socioeconomic status), employment status (working, retiring, or other), qualifications (with or without university/college degree), and ethnicity. Lifestyle behaviors included smoking status (current, previous smoker or not), alcohol consumption (current, previous drinker or not), alcohol consumption frequency (daily or almost daily, 3-4 times a week, 1-2 times a week, 1-3 times a month, special occasions only, and none), total physical activity (total metabolic equivalent task minutes per week), and healthy diet (greater intake of fruits, vegetables, whole grains, milk, fish or shellfish, dairy products, vegetable oils, and water, along with decreased or no intake of refined grains, processed and unprocessed meats, and sugar-sweetened beverages) (15). Social isolation was assessed by three aspects including number in household, friend/family visits, and leisure/social activities (16). For each aspect, low risk counted 0 score, and high risk counted 1. High-risk aspects were defined as living alone, friends/family visits occurring fewer times than once a month, and no participation in leisure/social activities. Social isolation was defined if the total social isolation score ≥ 2. We also gathered information on vitamin supplements, overall health status, C-reactive protein (CRP) level, and medication use ( Supplementary Table S3 ).

All participants were followed up from the date of recruitment until the date of disease diagnosis, death, or the end of the study period (December 17, 2022). Baseline characteristics of the study participants were summarized as mean with SD or median (interquartile range) for continuous variables and frequency and percentage for categorical variables. Cox proportional hazard models were used to estimate the hazard ratios (HRs) and 95% confidence interval (CI) for incident endocrine diseases associated with gastrointestinal diseases or the number of gastrointestinal comorbidities. As shown in supplement Supplementary Figure S1 , the design and analysis were divided into two parts. Firstly, we constructed a cohort for the baseline gastrointestinal diseases (excluding the baseline endocrine diseases) and performed Cox regressions for the risk of future endocrine diseases. Then, for the reverse association, we constructed a cohort for the baseline endocrine diseases (excluding the baseline gastrointestinal diseases) and performed another Cox regression for the risk of future gastrointestinal diseases. The proportional hazards assumption was tested by creating a time-dependent variable and no violation was found. In above Cox regression models, we adjusted for the baseline age (continuous), sex (male or female), ethnicity (White people, Black people, Asian, and other), BMI (continuous), Townsend deprivation index (continuous), employment status (working, retired, or other), qualifications (college attendance or not), physical activity (continuous), smoking status (current, previous, or never), alcohol consumption (current, previous, or never), adherence to a healthy diet (yes or no), social isolation (yes or no), and CRP (continuous).

A parallel analysis was conducted to evaluate the association of endocrine diseases or the number of endocrine comorbidities with the risk of gastrointestinal diseases, and vice versa.

In sensitivity analysis, to investigate whether the bidirectional association between gastrointestinal diseases and endocrine diseases differed by subgroups, stratification analyses were further performed for the main analyses according to sex (men or women), age (<60y or ≥60y), smoking status (current smokers or not), drinking status (current drinkers or not), BMI (obese or not), and medicine use (yes or no). To assess whether the identified relationships change over time, sensitivity analysis was conducted by excluding the participants who developed within 1 year of follow-up. To confirm the accuracy of our results, sensitivity analysis was further conducted by excluding participants who had missing data on covariates or participants with new-onset gastrointestinal diseases (as exposure group) or endocrine diseases (as exposure group) during follow-up. All analyses were performed using R (version 4.2.2) software. The statistical tests were two-sided, and the significance threshold was 0.05.

Table 1 presents the baseline characteristics of 481841 participants stratified by gastrointestinal disease status in detail. Among those, the mean age was 56.41 (SD: 8.10) years; 54% of participants were women. Compared with participants without gastrointestinal diseases, participants with gastrointestinal diseases were more likely to be smokers, have lower socioeconomic status, lower physical activity, and higher social isolation. They were less likely to work and have a college or university education. Supplementary Table S4 presents the baseline characteristics of 452858 participants stratified by endocrine disease status in detail.

During a median follow-up of 13.7 years (interquartile range 12.8, 14.5), we documented 8178 cases of endocrine diseases in participants with gastrointestinal diseases and 53190 cases in those without gastrointestinal diseases. Overall, gastrointestinal diseases were associated with total endocrine disease risk after multivariable adjustment (model 2: HR, 1.22; 95% CI, 1.19-1.25) ( Figure 1 ). Further adjustment for medication use did not substantially alter the results (HR, 1.14; 95% CI, 1.11-1.16) ( Supplementary Table S5 ). For individual endocrine diseases, participants with any gastrointestinal diseases had higher risks of hypothyroidism (HR, 1.17; 95% CI, 1.13-1.21), hyperthyroidism (HR, 1.54; 95% CI, 1.41-1.68), T2D (HR, 1.23; 95% CI, 1.20-1.27), hyperparathyroidism (HR, 1.48; 95% CI, 1.32-1.66), hyperpituitarism (HR, 1.40; 95% CI, 1.16-1.69), hypopituitarism (HR, 1.52; 95% CI, 1.22-1.90), ovarian dysfunction (HR, 1.84; 95% CI, 1.34-2.53), and testicular dysfunction (HR, 1.69; 95% CI, 1.26-2.28). Further adjustment for medication use did not substantially alter these results ( Supplementary Table S5 ).

In addition, we assessed the effect of individual gastrointestinal disease on the risk of total endocrine diseases. Duodenal ulcer (HR, 1.21; 95% CI, 1.10-1.34), gastritis and duodenitis (HR, 1.20; 95% CI, 1.16-1.25), dyspepsia (HR, 1.16; 95% CI, 1.10-1.23), irritable bowel syndrome (HR, 1.31; 95% CI, 1.21-1.43), intestinal malabsorption (HR, 1.29; 95% CI, 1.12-1.50), and gastrointestinal hemorrhage (HR, 1.19; 95% CI, 1.10-1.29) were significantly associated with an increased risk of total endocrine diseases ( Supplementary Figure S2 ).

During a median follow-up of 13.4 years (interquartile range 10.5, 14.3), we documented 5609 cases of gastrointestinal diseases in participants with endocrine diseases and 92369 cases in those without endocrine diseases. Overall, endocrine diseases were associated with total gastrointestinal disease risk after multivariable adjustment (model 2: HR, 1.48; 95% CI, 1.44-1.53) ( Figure 2 ). Further adjustment for medication use did not substantially alter the results (HR, 1.34; 95% CI, 1.30-1.38) ( Supplementary Table S6 ). For individual gastrointestinal diseases, participants with any endocrine diseases had higher risks of gastric ulcer (HR, 1.55; 95% CI, 1.42-1.70), duodenal ulcer (HR, 1.46; 95% CI, 1.29-1.65), gastritis and duodenitis (HR, 1.58; 95% CI, 1.52-1.64), dyspepsia (HR, 1.24; 95% CI, 1.14-1.35), irritable bowel syndrome (HR, 1.39; 95% CI, 1.29-1.51), intestinal malabsorption (HR, 1.79; 95% CI, 1.55-2.07), and gastrointestinal hemorrhage (HR, 1.43; 95% CI, 1.35-1.51). Further adjustment for medication use did not substantially alter these results ( Supplementary Table S6 ).

In addition, we assessed the effect of individual endocrine disease on the risk of total gastrointestinal diseases. Hypothyroidism (HR, 1.46; 95% CI, 1.39-1.53), hyperthyroidism (HR, 1.29; 95% CI, 1.14-1.45), T2D (HR, 1.55; 95% CI, 1.49-1.51), hyperparathyroidism (HR, 1.77; 95% CI, 1.45-2.17), and hyperaldosteronism (HR, 2.31; 95% CI, 1.55-3.45) were significantly associated with an increased risk of total gastrointestinal diseases ( Supplementary Figure S3 ).

The site-specific analysis associating individual gastrointestinal diseases and individual endocrine diseases is shown in Figure 3 and Supplementary Tables S7 , S8 . There is a bidirectional association between type 2 diabetes and six gastrointestinal diseases (gastritis and duodenitis, irritable bowel syndrome, gastrointestinal hemorrhage, dyspepsia, duodenal ulcer, and gastric ulcer); hyperthyroidism and four gastrointestinal diseases (gastritis and duodenitis, irritable bowel syndrome, malabsorption, and ulcerative colitis); hypothyroidism and four gastrointestinal diseases (gastritis and duodenitis, irritable bowel syndrome, malabsorption, and dyspepsia); hyperparathyroidism and three gastrointestinal diseases (gastritis and duodenitis, gastrointestinal hemorrhage, and duodenal ulcer).

Table 2 shows the association between the number of gastrointestinal comorbidity and total and individual endocrine diseases. We observed a significant trend (P trend) for almost all endocrine diseases when subtyped by the number of gastrointestinal diseases, with the exception of thyroiditis, Cushing syndrome, and hyperaldosteronism. For most related endocrine diseases, participants with multiple gastrointestinal comorbidities (two or more) exhibited higher risk estimates compared to those with a single gastrointestinal disease. Supplementary Table S9 shows the association between the number of endocrine comorbidity and total and individual gastrointestinal diseases. We observed a significant trend (P trend) for all non-neoplastic gastrointestinal diseases but not for neoplastic gastrointestinal diseases when subtyped by the number of endocrine diseases. For non-neoplastic diseases, participants with multiple endocrine comorbidities (two or more) exhibited higher risk estimates compared to those with a single endocrine disease.

In the sensitivity analyses, when we excluded participants who had missing data on covariates, incident cases that occurred within the first years of follow-up, or participants with new-onset exposure diseases during follow-up, the results consistent with the main findings ( Supplementary Tables S10 - S15 ). Stratified analyses revealed that the main findings were not different by sex, age, smoking status, alcohol intake, BMI, and medication use ( Supplementary Tables S16 - S19 ).