Society: AGA

Introduction: We aimed to investigate the relationship between consumption of ultra-processed foods (UPFs) and: 1) active symptomatic disease and 2) intestinal inflammation in a cohort of adults living with inflammatory bowel disease (IBD).

Methods: Participant data (n=141) was used from the prospective Manitoba Living with IBD study. Food intake was assessed using the Harvard Food Frequency Questionnaire (FFQ). UPF consumption was determined at 1-year follow-up using the NOVA Classification system. The percentage of total energy consumption from UPFs (NOVA4) was calculated and divided into 3 tertiles. We used linear regression analysis to assess the association between UPFs (using low [T1] vs. high [T3] consumption of UPFs) and active disease using the IBD Symptom Inventory (IBDSI) score of >14 for CD and >13 UC; and intestinal inflammation as measured by a fecal calprotectin level of >250 ug/g.

Results: The mean number of episodes of active symptomatic disease over one year was significantly higher among persons with UC, but not CD, in the higher consumption of UPFs group (T3) compared to the lower consumption of UPF (T1) group (13.9 vs 5.8, p=0.037). There were no significant differences in the mean number of episodes of inflammation over one year for lower vs. upper consumption of UPF in either UC or CD persons. When adjusting for age, gender, disease type and disease duration, the number of episodes of active symptomatic disease was 8 times less for the lower consumption of UPF (T1) compared to the higher consumption of UPF (T3) (Beta= -8.42, p=0.01) among UC.

Conclusion: UPF consumption appears to be a predictor of active symptomatic disease. While it was not significantly associated with increased inflammation, the study limitation of having fewer points of measurement in the year for FCAL may have underpowered this finding. Reducing UPF consumption is an additional dietary strategy that can be suggested as a means of minimizing symptomatic disease among people living with IBD.

Background:
The pathogenesis of Crohn’s disease (CD) involves complex interactions between host genetics, environment, and the intestinal microbiome. Bile acids (BAs) can act as signaling molecules involved in host immune regulation, and potentially in CD pathogenesis. Primary BAs help absorb dietary fat and are bio transformed into secondary BAs by the gut microbiome. However, the relationship between BAs, dietary fat, and CD development is unknown. We aimed to investigate the relationship between CD onset, BAs, and dietary fat in a pre-disease setting, and evaluate the predictive performance of these factors on CD onset via machine learning models.

Methods: Healthy first-degree relatives of CD patients were recruited as part of the Crohn’s and Colitis Canada Genetic Environment Microbial (GEM) project. In a nested case-control sub-cohort, 87 subjects diagnosed with CD on follow up were matched 1:4 to subjects who remained unaffected, by age, sex, follow-up time, and geographic location (n=347). Baseline serum, urine, and stool BAs were measured with the combination of four ultrahigh Performance Liquid Chromatography-Tandem Mass Spectroscopy. We used recruitment food frequency questionnaire data, deriving dietary fat types with the 2015 StatsCan Nutrient database. We used Generalized Estimation Equations to explore relationships between BAs (n=92) and dietary fat (n=9), and conditional logistic regressions to identify associations between BAs and dietary fats with CD. Finally, we used a tree-based machine-learning algorithm (XGBoost) with 5-fold cross-validation to assess the predictive performance of BA and dietary fat on future onset of CD. Two-sided p<0.05 defined significance.

Results:
In total, 22 BAs associated with dietary fat (p<0.05). Next, alterations in serum-derived cholate, glycocholate glucuronide, and glycoursodeoxycholic acid sulfate, stool-derived lithocholate, urine-derived deoxycholic acid 12-sulfate, and monounsaturated fat intake were associated with increased odds of CD (p<0.05). For CD onset prediction, serum-derived BAs had the best predictive performance (mean AUC of 0.70 [95% CI: 0.63-0.76]), followed by stool derived BAs (mean AUC= 0.61 [0.50-0.71]), and urine derived BAs (mean AUC= 0.52 [0.43-0.61]). In contrast, dietary fats were not predictive of CD onset (mean AUC= 0.49 [0.39-0.62]).

Conclusions: Currently, serum-derived BAs better predict the risk of CD than stool or urine derived BA, while dietary fat is not predictive of CD risk. BAs may play a role in the pathogenesis of CD, years before diagnosis.

Funding
The Leona M. and Harry B. Helmsley Charitable Trust
Kenneth Croitoru received the Canada Research Chair in Inflammatory Bowel Diseases
The International Organization for the Study of Inflammatory Bowel Diseases (IOIBD)
Jingcheng Shao received the Data Science Institute Summer Undergraduate Data Science award

Cristian Hernandez Rocha

Appearances