EFFECT OF CO-MORBIDITIES IN CROHN'S DISEASE ASSOCIATED URINARY METABOLIC PROFILES

Introduction Distinct metabolic signatures have been detected in urine that differentiate Crohn’s disease (CD) from controls in multiple studies, with consistent discriminatory metabolites derived from bacteria and co-bacterial pathways (Williams, 2009. AJG). Multiple other diseases have also been found to affect the urinary metabolome, and many of these relate to changes in bacterial associated metabolites (Lu, 2013. Front. Med.). This study aimed to examine a real life cohort of CD patients, and so included patients with other co-morbidities, to examine if the same metabolite changes were present, and if these patients could be distinguished from controls despite the presence of co-morbidities. Methods Nuclear magnetic resonance (H1NMR) spectroscopy was used to acquire urinary metabolic data from 74 CD patients and 100 controls. 19 of the CD group and 48 of the controls had at least one significant co-morbidity (diabetes, asthma, hypertension). Multivariate analysis was performed using OPLSDA. Univariate analysis was also performed to assess whether bacterial associated metabolites, as demonstrated in previous studies (Williams, 2009), were significantly different in CD patients compared to controls. These metabolites were Hippurate, Alanine, Citrate, P-Cresol, Phenyacetylglutamine (PAGn), and Dimethylglycine (DMG). Results OPLSDA analysis showed statistically significant separation between CD patients and controls irrespective of the presence of comorbidities. Model 1 compares CD patients to healthy controls (H). Models 2 and 3 include CD patients with at least one other co-morbidity (CDWC), and Model 3 includes non-CD patients with another co-morbidity in the control group (C). Univariate analysis showed that the bacterial associated metabolites Hippurate, Citrate, P-Cresol, DMG, and PAGn changed with statistical significance between CD groups and controls irrespective of the presence of co-morbidities. Conclusions The pattern of change in discriminating metabolites appear to be preserved in models separating CD from controls when patients with co-morbidities are included in the models, and these groups can be significantly separated with multivariate analysis. It is likely the effect of microbial disturbance is still measurable with this technique in a real world cohort.