Sustained modulation of intestinal bacteria by exclusive enteral nutrition used to treat children with Crohn’s disease


Dr A. S. Day, Department of Gastroenterology, Sydney Children’s Hospital, High Street, Randwick, Sydney, NSW 2031, Australia.


Background  The use of exclusive enteral nutrition to treat paediatric Crohn’s disease (CD) is widely accepted, although the precise mechanism(s) of action remains speculative.

Aim  To investigate the changes to key intestinal bacterial groups of Eubacteria, Bacteroides, Clostridium coccoides, Clostridium leptum and Bifidobacteria, during and after exclusive enteral nutrition treatment for CD in paediatric patients and correlate these changes to disease activity and intestinal inflammation.

Methods  Stool was collected from six children at diagnosis of CD, during exclusive enteral nutrition and 4 months post-therapy, and from seven healthy control children. The diversity of bacteria was assessed by polymerase chain reaction-denaturing gradient gel electrophoresis with changes to bacterial diversity measured by Bray–Curtis similarity, intestinal inflammation assessed by faecal S100A12 and the disease activity assessed by PCDAI.

Results  A significantly greater change in intestinal bacterial composition was seen with exclusive enteral nutrition treatment compared with controls. Further, the intestinal bacteria remained altered 4 months following exclusive enteral nutrition completion. Changes in the composition of Bacteroides were associated with reduced disease activity and inflammation.

Conclusions  Exclusive enteral nutrition reduces bacterial diversity and initiates a sustained modulation of all predominant intestinal bacterial groups. Exclusive enteral nutrition may reduce inflammation through modulating intestinal Bacteroides species. The implications of these results for exclusive enteral nutrition therapy and CD pathogenesis should now be the subject of further investigation.


Inflammatory bowel disease (IBD) is a life-long debilitating disorder of unknown aetiology. It is accepted that three factors, genetic predisposition, environmental trigger(s) and the intestinal micro-biota, combine to initiate disease. Further, there is clear evidence that the intestinal micro-biota is an essential component for disease development.1 Evidence is provided by both animal and human studies, which have shown that genetically predisposed mice develop IBD-like symptoms when intestinal micro-biota is present but germ free mice fail to develop IBD-like symptoms2 and colitis in humans can be eliminated by the diversion of the faecal stream.3 While a number of specific micro-organisms, including invasive E. coli4 have been implicated in the pathogenesis of IBD,5 to date, there is no conclusive evidence to suggest IBD is the result of a single infectious agent.6 Alternatively, there is the suggestion that the intestinal micro-biota as a whole may operate to drive inflammation in IBD.7

Exclusive enteral nutrition (EEN) using a polymeric formula is an important advance in IBD therapy. EEN is as efficacious as corticosteroids at inducing remission in paediatric IBD but without the side effects,8 highlighting the benefits of this therapy. However, further widespread use of EEN may be hindered by poor familiarity and an incomplete knowledge of the precise mechanism(s) of action. Previously, we have shown that polymeric formula has direct anti-inflammatory effects on the intestinal epithelial cells and may reduce inflammation, in part, through this process.9 In addition, there is evidence that EEN modulates the intestinal microflora, suggesting inflammation may also be reduced by changes to the intestinal micro-biota.10, 11 However, these initial studies were limited and only described changes to a narrow selection of intestinal bacteria.

Although there may be great inter-individual variation, the intestinal micro-biota is estimated to consist of approximately 400 species.12 Utilizing the technique of polymerase chain reaction (PCR) amplification of the bacterial 16S rDNA gene followed by denaturing gradient gel electrophoresis (DGGE)13 in combination with primers for five of the core groups of bacteria in human faeces, Eubacteria, Bacteroides–Prevotella, Clostridium coccoides, Clostridium leptum and Bifidobacteria, greater than 90% of the predominant bacterial species in the human intestine can be identified.14–16 Furthermore, changes within these groups can be quantitated by applying Bray–Curtis similarity testing.17

In addition to investigating the intestinal micro-biota, an accurate measure of intestinal inflammation is also required. Faecal sampling is considered the most accurate, non-invasive measure of intestinal inflammation and several faecal markers, each with their own advantages and disadvantages, are now available. S100A12 is one such marker, which is expressed primarily in the cytoplasm of granulocytes and released rapidly during an inflammatory response.18 Further, the assay has been validated for children19 and faecal S100A12 has been shown to be elevated in paediatric IBD and can be used to readily distinguish IBD from non-organic intestinal disease.20 Therefore faecal S100A12 is an appropriate method of non-invasively assessing intestinal inflammation.

Therefore, our hypothesis was that EEN alters the intestinal microflora, which contributes to reducing intestinal inflammation in IBD. Our aim was to investigate the changes in a diverse spectrum of prominent intestinal bacterial species, during and after EEN therapy, in children with CD. And further, to correlate changes in the micro-biota with changes in disease activity and intestinal inflammation to elucidate the potential mechanism of action of EEN.

Materials and methods

Patients and controls

Children attending the Sydney Children’s Hospital (SCH), Randwick, IBD clinic were recruited prospectively at the time of diagnosis of CD. The diagnosis of CD was based upon histologic, endoscopic and radiologic criteria, after performing upper gastrointestinal endoscopy, colonoscopy and small bowel barium follow-through study. Further, these children had not been exposed to immunomodulatory agents or other IBD therapies that have the potential to modulate or modify the intestinal micro-biota. Following diagnosis, children were prescribed Osmolite (Abbott Laboratories, Cronulla, NSW, Australia), a polymeric formula containing whole polypeptides and all nutritional requirements, to be taken as their EEN source as per the standard protocol.21 The commencement of EEN was within 2 days of diagnosis. Patient demographics (gender and age) were collected at diagnosis. Exclusion criteria for enrolment were previous diagnosis of CD or UC, severe colitis requiring intensive medical or surgical management and use of antibiotics or anti-inflammatory agents in the previous 4 weeks.

Normal healthy children from families and staff at SCH, volunteered as controls. These children did not have a current diagnosis of gastrointestinal disease, did not display gastrointestinal symptoms, did not take antibiotics in the 4 weeks prior to sample collection and did not have a significant change in diet over the study period. Further, these children were not related to the children with CD.

Disease location and paediatric Crohn’s disease activity index

Disease location was based on endoscopic and histological findings and classified as L1 (terminal ileum), L2 (colon), L3 (ileocolon), L4 (upper GI) or a combination as outlined by the Montreal classification.22

The paediatric Crohn’s disease activity index (PCDAI) for each patient was calculated at diagnosis (baseline) and at 2, 4, 6, 8, 16 and 26 weeks postcommencement of EEN.23 Disease remission was defined as a PCDAI <15.21

Faecal sample collection

For the children diagnosed with CD, faecal samples were collected at baseline (prior to endoscopy) and then at 1, 2, 4, 6, 8, 16 and 26 weeks after diagnosis. A baseline faecal sample was collected from each of the control subjects along with a follow-up sample 6–8 weeks later. Samples were collected in sterile collection containers (Techno-Plas, St Marys, SA, Australia) and immediately stored in the home freezer at −20 °C, then transported frozen to the laboratory where they were stored at −80 °C until analysis.

Informed consent was obtained from each child or their care-giver(s). The project was approved by the South Eastern Sydney and Illawarra Area Health Service Research Ethics Committee.

DNA extraction from faecal samples

Frozen faecal samples were thawed to room temperature and an aliquot of approximately 200 mg stool was collected and weighed. DNA was extracted from the stool aliquot using the QIAamp DNA stool mini kit (Qiagen Pty Ltd, Germantown, MD, USA) according to the manufacturer’s instructions. This kit is specifically designed to remove compounds that may be present in human faeces that can degrade DNA and inhibit the PCR. Following extraction, the DNA concentration and purity was checked by spectrophotometry using a NanoDrop-1000 (Thermo Fisher, Wilmington, DE, USA). Only samples with a 260/280 nm ratio of 1.8–2.0 and a concentration greater than 5 ng/μL were used for further analysis. Samples that did not meet this criterion were re-extracted. If the re-extracted samples again failed to meet the above criterion, the samples were excluded from further analysis.


Bacterial 16S rDNA genes were amplified from the extracted faecal DNA using primers for five groups of bacteria: Eubacteria, Bacteroides–Prevotella, C. coccoides, C. leptum and Bifidobacteria. The primer sequences and accompanying references describing the bacterial species targeted by each primer set are shown in Table 1. A hot start PCR was used to amplify the 16S rDNA genes. Each 25 μL PCR contained 10 ng DNA template, reaction buffer (Fisher Biotech, Subiaco, WA, Australia), MgCl2 (Fisher Biotech), 2 mm of nucleotides (dNTP; Fisher Biotech), 10 pmol of each forward and reverse primer (Fisher Biotech) and 0.825 units Taq DNA polymerase (Fisher Biotech). The PCRs were run on a MJ Mini Personal Thermocycler (BioRad, Hercules, CA, USA). PCR tubes were heated at 94 °C for 5 min to facilitate the hot start, then run for 30 cycles each under the conditions stated in Table 1. To verify the suitability of the PCR products for DGGE, they were separated by agarose gel electrophoresis, stained with ethidium bromide and visualized with UV illumination (Gel Doc 2000; BioRad). Only PCR products that displayed a single band at the correct size were subject to further DGGE analysis.

Table 1.   Details of PCR-DGGE for the Eubacteria, Bifidobacteria, Bacteroides–Prevotella, Clostridium coccoides and Clostridium leptum bacterial groups
GroupEubacteriaBifidobacteriaBacteroides–PrevotellaClostridium coccoidesClostridium leptum
Product size (bp)466532399579538
PCR cycle
 Separation 94 °C 30 s94 °C 30 s94 °C 20 s94 °C 20 s94 °C 20 s
 Annealing 56 °C 20 s 62 °C 20 s 64 °C 30 s 63 °C 30 s 63 °C 30 s
 Elongation 72 °C 60 s 72 °C 60 s72 °C 60 s72 °C 60 s72 °C 60 s
Denaturing gradient (%)43–5743–6540–7043–6038–65

Six per cent polyacrylamide gels containing a gradient of denaturant (100% denaturant is defined as 7 m urea and 40% deionized formamide) were prepared for each individual primer set as indicated in Table 1. Approximately 15 μL of PCR product was loaded onto the gel. Samples from the same patient/control were grouped together to facilitate interpretation of the DGGE results. The electrophoresis was performed in 1X TAE buffer (0.2 m Tris[hydroxymethyl]aminomethane, 0.1 m acetic acid and 0.05 m ethylenediaminetetraacetic acid) for 16 h at 75 V and 60 °C (DCode System; BioRad). Following electrophoresis, gels were stained with ethidium bromide for 5 min followed by 15 min of destaining in MilliQ water and then viewed under UV illumination (BioRad).

Bray–Curtis analysis of percentage similarity

The banding profiles of the DGGEs were converted into binary arrays (1 indicating the presence of a band and 0 indicating the absence of a band). Binary arrays were then entered into the Plymouth Routines in Multivariate Ecological Research software program (PRIMER 5, Version 5.2.2; Primer-E Ltd, Plymouth, UK) and Bray–Curtis similarity for binary data was used to calculate the percentage similarity for repeat samples collected from the same individual. An example of calculating the percentage similarity is if the DGGE banding profiles of two samples (taken from the same individual at different time points) each display 10 bands. If each of the 10 bands is in-line with a corresponding band on the adjacent profile, then the samples are 100% similar, indicating the bacterial composition is the same. However, if only of the five bands are in-line between the two profiles and the other five bands are at alternating positions that do not line up, then these samples have a Bray–Curtis similarity of 50%. This indicates 50% of bacteria is common to both samples and 50% of bacteria is unique to each sample.

Measurement of faecal S100A12

Faecal S100A12 levels were measured by immunoassay following an extraction process. The faecal extraction process was derived from the PhiCal test product insert (Nycomed Pharma AS, Oslo, Norway) and is summarized as follows. Approximately 100 mg of faecal material was obtained from frozen stool samples. To minimize thawing, the extraction buffer (Nycomed) was added at a dilution of 1:50 as quickly as possible after which the faecal material in the extraction buffer was vortexed briefly with an agitator, before being placed on a horizontal platform mixer for 25 min at 120 rpm. Approximately 1.4 mL of homogenate was then transferred to a new tube and centrifuged at 10 000 g for 20 min. One millilitre of clear supernatant was then transferred to a new tube and stored at −80 °C until the assay was performed. S100A12 levels in the faecal extracts were measured by an in-house ELISA as previously described.19


All statistics and graphs were generated with GraphPad Prism version 4.00 for Windows (GraphPad Software, San Deigo, CA, USA). DGGE band counts and percentage similarity are presented as the mean with standard deviation. The difference in band count between the CD and controls, at baseline and week 8, was assessed by anova with Bonferroni post hoc test. Differences between control and CD percentage similarity were assessed by t-test. Differences between CD and control faecal S100A12 levels at baseline were assessed by the Mann–Whitney test. Spearman correlation was used to correlate PCDAI and Faecal S100A12. Pearson correlation was used for all other correlations. PCDAI and S100A12 results are presented as the median and range. Significance was accepted if P < 0.05.


Patient demographics

Six children with CD and seven control children were recruited (Table 2). Children with CD were treated with EEN for a median of 56 days (range 46–57 days). Following EEN, five of the six children continued with supplementary EN, whilst two of the six began aminosalicylate therapy and two of the six started azathioprine. Stools from the controls were collected at a median of 59 days apart (range 42–64 days).

Table 2.   Demographics of children with Crohn’s disease and control children
  1. CD, Crohn’s disease.

Age (years), mean (s.d.)5.9 (3.6)10.2 (4.5)
Age range (years)2.1–122.5–13.5
Gender, M/F3/44/2

Disease location, disease activity (PCDAI) and faecal S100A12

Disease location information was unavailable for one CD patient. Of the remaining five, all had L4 (upper GI) involvement with 1/5 L1 (terminal ileum), 3/5 L2 (colonic) and 1/5 L3 (ileocolonic) involvement.

At diagnosis (baseline), the children with CD had a median disease activity (PCDAI) of 26.25 (range 10–37.5), which decreased to a median of 17.5 (range 0–22.5) at the completion of EEN (Figure 1). Disease activity was further reduced at 26 weeks with all but one child in remission (PCDAI <15; median 3.75, range 0–17.5; Figure 1).

Figure 1.

 Disease activity [paediatric Crohn’s disease activity index (PCDAI)] and intestinal inflammation (faecal S100A12) in children with Crohn’s disease. The PCDAI (white box plot) was calculated at baseline (week 0) and for weeks 2, 4, 6, 8, 16 and 26 postcommencement of exclusive enteral nutrition (EEN). There is a significant reduction (P < 0.05) in PCDAI at week 26 compared with baseline. Faecal S100A12 (grey box plot) was also calculated at the time points of baseline and for weeks 2, 4, 6, 8, 16 and 26 postcommencement of EEN.

Faecal S100A12 levels were measured in all samples to provide an indication of the intestinal inflammation. The median faecal S100A12 level at baseline in children with CD was 59.9 mg/kg (range 6.2–279.5 mg/kg; Figure 1), which was significantly higher (P < 0.01) than the faecal S100A12 levels in the control samples (median 6.5, range 3.9–17.3 mg/kg). For children with CD, the median faecal S100A12 levels at weeks 8 and 26 were 61.4 mg/kg (range 10.7–347.8 mg/kg) and 39.1 mg/kg (range 4.9–299.6 mg/kg) respectively (Figure 1). A significant correlation was found between faecal S100A12 levels and PCDAI (R = 0.5299, P = 0.0018).

Band count analysis of PCR-DGGE

All stools collected from children with CD and the controls were subjected to DNA extraction and PCR-DGGE. A DGGE profile with more bands generally indicates a greater degree of bacterial species diversity compared with a DGGE profile with few bands. Therefore, the number of bands that appeared on the DGGE was quantitated and used to estimate bacterial diversity in each sample. To facilitate the analysis, band counts were averaged for each group. Band counts for children with CD were compared with controls for the baseline and week 8 time points. The Eubacteria results indicated that there was no significant difference in band count at baseline between CD and controls; however, the band count for children with CD at week 8 was significantly lower compared with controls (anovaP < 0.05, post hoc P < 0.05) (Figure 2a). The Bifidobacteria and C. leptum PCR-DGGE results indicated that there was no difference in band count for these two primer sets at either the baseline or week 8 time point (anovaP > 0.05; Figure 2b,e). However, the Bacteroides–Prevotella (anovaP < 0.01, post hoc P < 0.05) and C. coccoides (anovaP < 0.001, post hoc P < 0.001) band counts were significantly decreased in the CD group compared with the control at week 8 (Figure 2c,d).

Figure 2.

 PCR-DGGE band counts. (a) Eubacteria, (b) Bifidobacteria, (c) Bacteroides–Prevotella, (d) Clostridium coccoides and (e) Clostridium leptum PCR-DGGE was performed on seven control children (broken line; weeks 0, 8) and six children with Crohn’s disease (CD; weeks 0, 1, 2, 4, 6, 8, 16, 26). Mean and standard deviation counts are presented for each time point. Band count was significantly different (P < 0.05) between CD and control for Eubacteria at week 8, Bacteroides–Prevotella at weeks 0 and 8 and Clostridium leptum at week 8. DGGE, denaturing gradient gel electrophoresis.

Bray–Curtis similarity analysis of PCR-DGGE

To further analyse the stool bacterial composition, Bray–Curtis similarity testing was performed on the PCR-DGGE results to investigate the degree of change in bacterial diversity within an individual over time. For the controls, the similarity in bacterial composition in the two samples taken 8 weeks apart was high (Table 3) indicating a relatively stable composition in stool bacteria over this time. The most stable of the bacterial groups was Bifidobacteria, which remained 84% similar over the 8 weeks and the least stable being the C. leptum group, which retained only 59% similarity over the 8 weeks (Table 3). In comparison, during the 8 weeks of EEN treatment, children with CD had a greater degree of change in the bacterial composition for all the bacterial groups examined (Table 3). The most stable group was Bacteroides–Prevotella, which still only remained 38% similar following EEN treatment compared with the composition prior to the treatment (Table 3). Similar to the changes observed in the controls over the 8-week period, the least stable group following EEN treatment was C. leptum, which retained 15% similarity following EEN treatment (Table 3). Statistical comparison indicated that for all the bacterial groups, this change in similarity for children with CD treated with EEN was significantly greater than the change in similarity for the healthy controls, who were on a normal diet, over the same time period (Table 3).

Table 3.   Percentage similarity of baseline (week 0) compared to 8-week stool samples collected from controls and children with Crohn’s disease treated with exclusive enteral nutrition
ComparisonControlCD (EEN treatment)P-value
0–8 weeks0–8 weeks
Mean (%)s.d.nMean (%)s.d.n
  1. CD, Crohn’s disease; EEN, exclusive enteral nutrition.

Bacteroides 8313738225<0.01
Clostridium coccoides7413728356<0.01
Clostridium leptum5930715216<0.05

We were also interested in investigating the changes to the intestinal micro-flora following completion of the EEN therapy. Stools were collected from children with CD at two further time points: 8 weeks following completion of EEN (16 weeks postdiagnosis) and 4 months following completion of EEN (26 weeks postdiagnosis). Further changes in the intestinal bacterial composition were observed following the completion of EEN. The greatest change appeared to occur immediately following completion of EEN from weeks 8 to 16. For example, the PCR-DGGE banding profiles for Eubacteria and C. coccodies were, respectively, 14% and 16% similar, comparing the week 8 sample with the week 16 sample, whereas they were, respectively, 48% similar and 37% similar comparing the week 16 sample with the week 26 sample (Table 4). However, when the similarities at weeks 26–0 were compared, all the bacterial groups appear to have changed in composition by approximately the same amount. That is, all the bacterial groups were between 31–41% similar in composition for the week 26 sample compared to the baseline (week 0) sample (Table 4).

Table 4.   Percentage similarity of baseline, weeks 8, 16 and 26 stool samples collected from children with Crohn’s disease
Comparison0–26 weeks8–16 weeks16–26 weeks
Mean (%)s.d.nMean (%)s.d.nMean (%)s.d.n
Bacteroides 336650NA10NA1
Clostridium coccoides363531627340293
Clostridium leptum41226712337333

Correlation of changes in the intestinal micro-biota to disease activity and inflammation

The changes in intestinal micro-biota, as measured by Bray–Curtis % similarity, were correlated to changes in disease activity (ΔPCDAI) and inflammation (ΔS100A12) for the following time periods: weeks 0–8, 8–16, 16–26, 8–26 and 0–26. Significant correlations were found for the weeks 0–8 and 8–26 time periods only. For weeks 0–8, ΔPCDAI correlated to Bacteroides–Prevotella% similarity (R2 = 0.8834, P = 0.0175; positive slope; Figure 3) and ΔS100A12 correlated with Eubacteria % similarity (R2 = 0.7384, P = 0.0283; negative slope). For weeks 8–26, ΔPCDAI correlated with C. leptum% similarity (R2 = 0.8063, P = 0.0385; negative slope) and ΔS100A12 also correlated with C. leptum% similarity (R2 = 0.7716, P = 0.0500; negative slope).

Figure 3.

 Modulation of Bacteroides–Prevotella group bacteria during exclusive enteral nutrition (EEN) correlates with change in paediatric Crohn’s disease activity index (PCDAI). Stool samples from children with Crohn’s disease were collected prior to (week 0) and at completions of (week 8) EEN and subjected to PCR-DGGE and similarity analysis. A positive correlation (R2 = 0.7384, P = 0.0283) was found between percentage similarity of Bacteroides–Prevotella group bacteria and change in PCDAI (delta PCDAI) for the period of EEN treatment.


This study is the first to use a diverse array of primer sets (Eubacteria, Bifidobacteria, Bacteroides–Prevotella, C. coccoides and C. leptum) to investigate in detail the intestinal micro-biota response to EEN treatment, and to correlate these changes to disease activity and intestinal inflammation. This study has also investigated the longer-term (4-month) changes to the intestinal micro-biota postcompletion of EEN treatment.

The results of the current study show that at diagnosis, bacterial diversity in children with CD was similar to controls. This is in contrast to a report from Manichanh et al.24 who reported a reduced diversity in Firmicutes (specifically the C. coccoides and C. leptum groups) in adult patients with quiescent CD. These contrasting findings may be a result of differences in the study populations. That is, Manichanh et al. investigated adults with pre-existing but quiescent CD, as compared with the current study, which investigated children with active newly diagnosed CD who had not been previously treated. Nevertheless, changes in diversity in the Eubacteria, Bacteroides–Prevotella and C. coccoides groups following EEN treatment detected in the current study indicates that EEN does alter intestinal bacterial diversity.

Changes to bacterial diversity may represent a possible mechanism of action of EEN. Therefore, similarity analysis was used in an attempt to quantitate changes to the intestinal bacterial composition with EEN. The results of the current study concurs with previous findings that EEN does modulate intestinal bacteria;10, 11 however, in the current study, the degree of change during EEN is significantly greater than the change observed in healthy children on a normal diet over the same period of time. In addition, the current study more precisely defined these changes. That is, with 8 weeks of EEN, the greatest change in bacterial composition was seen in C. leptum group bacteria and the least change was seen in Bacteroides–Prevotella group bacteria. However, all bacterial groups had a greater degree of change with EEN than in controls.

To elucidate whether changes to these specific bacterial groups may be a mechanism by which EEN reduces intestinal inflammation, we correlated changes in the individual bacterial groups with changes in disease activity and faecal S100A12. Of interest was that with EEN treatment, we found a significant positive correlation between Bacteroides–Prevotella and PCDAI. This indicates that when there was a large change to Bacteroides–Prevotella group bacteria, there was a large reduction in PCDAI. The small sample size does limit our ability to draw firm conclusions from these interesting results. However, previous reports have also implicated Bacteroides–Prevotella species as being associated with the inflammatory process in IBD.25, 26 Therefore, together these results suggest that Bacteroides–Prevotella may have a role in the disease process and further investigation of this aspect of disease pathogenesis should be pursued.

We also investigated what happened to the intestinal bacterial composition after the completion of EEN. Post-therapy all children with CD returned to a solid food diet with the majority continuing with supplementary EN and with the addition of further medications (aminosalicylate and azathioprine). Two months following EEN therapy the intestinal bacterial composition continued to change, and intestinal bacterial composition was again different at 4 months post-therapy. When samples of 4 months post-therapy were compared with samples from diagnosis, all five bacterial groups were approximately 40% similar. Therefore, EEN has an immediate and pronounced effect on the intestinal microbial composition, In addition, 4 months following EEN therapy, there is a partial, but not complete, reversion back to the bacterial profile at baseline. Further, we found that changes to C. leptum group bacteria negatively correlated with both changes in PCDAI and faecal S100A12 suggesting that post-EEN, stability in the C. leptum group composition is associated with a reduced intestinal inflammation and disease activity. This is an interesting result and suggests that C. leptum probiotic therapy may be beneficial post-EEN therapy. However, more thorough investigation of this preliminary finding is required.

This study has produced a number of interesting results that may have potential therapeutic implications for the management of IBD, however caution must be taken when interrupting these results. The intensive nature of the analysis and the large number of samples included for each study subject have limited the sample size of this study. Nevertheless, this study has identified areas of specific interest that can now be investigated with targeted studies and larger sample sizes, to confirm these findings.

In summary, this study has shown for the first time that EEN has a significant and sustained effect on the composition of the predominant intestinal bacterial groups. And further, this study has also identified changes in Bacteroides–Prevotella group bacteria during EEN treatment that are associated with reduced disease activity. The implications of these findings to IBD pathogenesis and therapy must now be the subject of further investigation.


Declaration of personal interests: None. Declaration of funding interests: This work was supported in part by a UNSW Faculty Research Grant (ASD and HMM). Laboratory investigations were conducted in the Westfield Research Laboratories. Osmolite provided by Abbott Australasia.