P804 Development and Validation of the Glasgow Exclusive Enteral Nutrition Index of Compliance (GENIE)

Abstract

Abstract Background Treatment adherence is key to the efficacy of exclusive enteral nutrition (100% EN) in active Crohn’s disease (CD), but there are currently no biomarkers to objectively support this. We explored faecal parameters as putative biomarkers of compliance with 100% EN, and subsequently developed and validated the Glasgow Exclusive enteral Nutrition Index of compliancE (GENIE). Methods Healthy adults replaced all (100% EN) or part (85% EN, 50% EN and 25% EN) of their habitual diet with a polymeric formula (Modulen IBD, Nestle©) for 7 days. Faecal pH, Bristol stool chart score, water content, short chain fatty acids (SCFAs) and branched chain fatty acids (BCFAs) were measured before (D0) and after (D7) each intervention. Faecal biomarkers and threshold values for group assignments were derived using receiver operating characteristic (ROC) curve analyses and machine learning algorithm to develop the GENIE. The GENIE was validated in 30 children with CD, during 100% EN and 4 weeks after return to normal diet. Results Sixty-one (31 females) adults (mean age [SD]: 25.9 [4.3] years) were recruited. D7 faecal pH and the ratios of BCFAs to either acetate or butyrate performed the best to differentiate between patients on 100% EN from <100% EN. A ratio of isobutyrate (IC4) to C2> 0.039 and a ratio of IC4 to C2+C4> 0.035 both produced a sensitivity, specificity, and positive predictive value (PPV) of 92%, 83% and 79%, respectively to differentiate between 100% EN vs. all other groups. A faecal pH> 8.0 produced a sensitivity, specificity and PPV of 84%, 86% and 81%, respectively. Findings using machine learning were better than those using ROC curve analysis. Two models were generated; one which includes all faecal metabolites (Laboratory GENIE, L-GENIE), and which produced a sensitivity, specificity and positive predictive value (PPV) of 88%, 94%, and 92%, respectively and a second one (Clinical GENIE, C-GENIE) which considers only faecal pH as the input data and which produced a sensitivity, specificity and PPV of 84%, 86% and 81% (Figure 1). Validation of GENIE models in children with CD found that C-GENIE outperformed L-GENIE, producing a sensitivity, specificity and PPV of 85%, 88% and 88%, respectively (Table 1). Figure 1: Glasgow Exclusive Enteral Nutrition Index of Compliance. Table 1: Results of validation of the GENIE model in children with Crohn’s disease (CD) during treatment with 100% EN and after return to unrestricted diet and compared to the development cohort of healthy adults. Conclusion GENIE can predict adherence to 100% EN in patients with CD in both the clinical and research settings. It may complement conventional dietary assessment and inform clinical decision, especially in patients who are not in remission.