P518 Dietary and non-dietary predictors of treatment response to adalimumab in anti-TNFα-naïve adults with Crohn’s disease

Abstract

Background Biologics, such as anti-TNFα agents, are commonly used in the management of Crohn’s disease (CD). A significant proportion of patients do not respond to treatment, necessitating the exploration of pre-treatment predictors of treatment response. Methods Anti-TNFα-naïve adults with active CD (Crohn’s Disease Activity Index; CDAI≥150) participating in an RCT (NCT04859088) were randomised to receive adalimumab monotherapy or adalimumab combination therapy with 50% partial enteral nutrition (PEN). Treatment response (CDAI<150) was assessed after 6 weeks, baseline diet was assessed with EPIC-Norfolk FFQ, alternative Mediterranean diet scores (aMED), and principal component analysis (PCA) with orthogonal (varimax) rotation was used to identify data-derived dietary patterns. Baseline predictors evaluated included PEN use, steroid use, immunomodulator use, age, disease duration, CDAI, C-Reactive protein (CRP), albumin, haemoglobin, Scottish Index of Multiple Deprivation (SIMD) score, adherence to dietary patterns identified, aMED score, smoking status, alcohol consumption, physical activity level, body mass index (BMI), fat mass (kg/m2), fat-free mass (kg/m2), and handgrip strength. Differential analysis between responders and non-responders was carried out with general linear model or chi-square test when appropriate. Random forest model with recursive feature elimination (RF-RFE) was used to identify the most predictive factors of treatment response. Results Of 42 participants recruited to the study, 62% (26) responded to treatment. PCA revealed four dietary patterns (Figure 1A). Responders to adalimumab were younger (mean (SD): 36.0 (17.1) vs 50.8 (10.0), P=0.004), had lower baseline CDAI (mean (SD): 228 (62) vs 286 (78), P=0.018), higher CRP (14.5 (19.2) vs 4.6 (5.8) mg/L, P=0.036), were less likely to smoke (31% (5 of 16) vs 8% (2 of 26), and less likely to adhere to a dietary pattern characterised by high consumption of animal products (PC2) (P=0.030). Adherence to PC2 also correlated positively with age (r=0.327, P=0.035). The RF-RFE algorithm highlighted young age, low baseline CDAI and low PC2 adherence as key factors (Sensitivity: 77%, Specificity: 63%, PPV: 77%, NPV: 63%, OOB: 29%, P=0.012) (Figure 1B). Interestingly, exclusion of dietary factors improved diagnostic performance of the model (Sensitivity: 77%, Specificity: 75%, PPV: 83%, NPV: 67%, OOB: 24%, P=0.003) (Figure 1C), indicating potential interactions by other factors like age. Conclusion Young age, non-smoking, low baseline CDAI and elevated CRP predict adalimumab response in anti-TNFα-naïve adults. While dietary factors may also play a role, their impact seems confounded by other non-dietary factors. Further research is warranted in this area.