Pre-Diagnosis Diet Predicts Response to Exclusive Enteral Nutrition and Correlates with Microbiome in Pediatric Crohn Disease

Abstract

Exclusive enteral nutrition (EEN) is effective in inducing remission in pediatric Crohn disease (CD). EEN alters the intestinal microbiome, but precise mechanisms are unknown. We hypothesized that pre-diagnosis diet establishes a baseline gut microbiome, which then mediates response to EEN. We analyzed prospectively recorded food frequency questionnaires (FFQs) for pre-diagnosis dietary patterns. Fecal microbiota were sequenced (16SrRNA) at baseline and through an 18-month follow-up period. Dietary patterns, Mediterranean diet adherence, and stool microbiota were associated with EEN treatment outcomes, disease flare, need for anti-tumor necrosis factor (TNF)-α therapy, and long-term clinical outcomes. Ninety-eight patients were included. Baseline disease severity and microbiota were associated with diet. Four dietary patterns were identified by FFQs; a "mature diet" high in fruits, vegetables, and fish was linked to increased baseline microbial diversity, which was associated with fewer disease flares (p < 0.05) and a trend towards a delayed need for anti-TNF therapy (p = 0.086). Baseline stool microbial taxa were increased (Blautia and Faecalibacterium) or decreased (Ruminococcus gnavus group) with the mature diet compared to other diets. Surprisingly, a "pre-packaged" dietary pattern (rich in processed foods) was associated with delayed flares in males (p < 0.05). Long-term pre-diagnosis diet was associated with outcomes of EEN therapy in pediatric CD; diet-microbiota and microbiota-outcome associations may mediate this relationship.

Keywords: dietary pattern; inflammatory bowel diseases; microbiome; nutrition; pediatrics; prediction.

Figure 1

Patient characteristics and major outcomes. (A): Flowchart of patient numbers throughout inclusion and outcome criteria. After exclusion, 98 patients were included in the analysis, of whom 76 did not require steroids, but 48 of those were initiated on anti-TNF therapy during maintenance. EEN = exclusive enteral nutrition, 5ASA = 5-aminosalicylate, anti-TNF = anti-tumor necrosis factor-α, GCS = glucocorticoid. (B): Baseline clinical characteristics of 103 patients included in the study. (C): Physician global assessment (PGA) was available for 48 patients at baseline and 6 months; PGA improved from baseline to 6 months follow-up. (D): Fecal calprotectin (FCP) tended to decrease from baseline to 2 months for the eight patient responders with paired specimens. **** = p < 0.0001.

Figure 2

Dietary associations with baseline and long-term clinical outcomes. (A): Mature diet adherence was negatively associated with SES-CD score at baseline in females. (rho = −0.4556, p < 0.05, n = 34 with relevant data available). (B): Mature diet adherence was negatively associated with baseline FCP in males (rho = −0.4556, p < 0.05, n = 18). (C,D): rMED score was negatively associated with SES-CD score at baseline and showed a smaller reduction in PGA in females. (rho = −0.42, p < 0.05, n = 34, rho = −0.4406, p < 0.01, n = 36, respectively). (E): Higher meat adherence was associated with a loss of response to EEN in males (p < 0.05). (F): A higher rMED score was associated with a decreased likelihood to show improvement in PGA from baseline to six months. (G): Lower adherence to a pre-packaged diet was associated with earlier development of disease flare in males, p < 0.05, n = 59. (H): Lower adherence to a mature diet was associated with an earlier need for anti-TNF, p = 0.086, n = 76. * = p < 0.05, tertile 1 = lowest adherence, tertile 3 = highest.

Figure 3

Dietary patterns and rMED score are associated with baseline microbial abundances, microbial diversity, and richness. (A,B): Baseline Blautia ASV1 and Faecalibacterium ASV2 relative abundances were positively associated with mature diet adherence (rho = 0.3989, p < 0.05, n = 35; rho = 0.4801, p < 0.01, n = 35). Blautia ASV2 and Faecalibacterium ASV1 were not significantly associated (p > 0.05). (C): Baseline Ruminococcus gnavus group ASV relative abundance was negatively associated with mature diet adherence (rho = −0.4179, p < 0.05, n = 35). (D): Baseline Blautia relative abundances were positively associated with rMED score (ASV1 rho = 0.4004, p < 0.05, n = 36; ASV2 rho = 0.4635, p < 0.01, n = 36). (E): Faecalibacterium ASV1 baseline relative abundance was positively associated with rMED score at baseline (rho = 0.3273, p = 0.0514, n = 36), ASV2 baseline relative abundance was positively associated with rMED score (rho = 0.4322, p < 0.01, n = 36). (F): Ruminococcus gnavus group ASV baseline relative abundance was negatively associated with rMED score (rho = −0.4444, p < 0.01, n = 36). (G–I): Simpson diversity at baseline was not associated with mature diet adherence, but mature diet adherence was associated with increased Shannon diversity and richness (Chao1) at baseline (rho = 0.2787, p > 0.05, n = 35; rho = 0.3989, p < 0.05, n = 35; rho = 0.4968, p < 0.01, n = 35). (J–L): Increased vegetarian diet adherence was associated with increased Simpson and trending Shannon microbial diversity, but not richness (Chao1), at baseline (rho = 0.3583, p < 0.05, n = 35; rho = 0.3323, p = 0.0508, n = 35; rho = 0.1347, p > 0.05, n = 35). (M–O): rMED scores were positively associated with Simpson and Shannon diversity as well as microbial richness (Chao1) at baseline (rho = 0.3302, p < 0.05, n = 36; rho = 0.4454, p < 0.01, n = 36; rho = 0.5218, p < 0.01, n = 36).

Figure 4

Pre-diagnosis diet is associated with microbial diversity at 6 months. (A–C): Increased mature diet adherence was associated with increased Simpson diversity, increased Shannon diversity, and an increase in richness at 6 months (rho = 0.4234, p = 0.0559, n = 24; rho = 0.4675, p < 0.05, n = 24; rho = 0.3891, p = 0.0813, n = 24). (D–F): Pre-packaged diet adherence was not associated with Simpson diversity at 6 months (rho = −0.2945, p > 0.05) but was inversely associated with Shannon diversity and richness at 6 months (rho = 0.3984, p < 0.05, n = 23; rho = 0.5059, p < 0.01, n = 23).

Figure 5

Lower baseline microbial diversity is associated with decreased time to disease flare. A lower Simpson diversity at baseline was associated with decreased time until disease flare requiring GCS; tertiles are shown for illustrative purposes; n = 29, with three disease flares. Tertile 1 = lowest diversity, tertile 3 = highest.

Figure 6

Linking pre-diagnosis diet to treatment outcomes through fecal microbiota. Our study revealed multiple links between pre-diagnosis diet, microbes, and patient outcomes, which are summarized in this figure. Mature and vegetarian diets and rMED score were positively associated with baseline microbial diversity, while pre-packaged diet was negatively associated with six-month diversity. Increased baseline diversity was associated with prolonged survival until disease flare. A decreased pre-packaged diet adherence and increased meat diet adherence were associated with decreased time until disease flare and lack of EEN response, respectively. Finally, increased rMED score was associated with a worse six-month PGA and a decreased likelihood for PGA improvement from baseline.