2.3. Statistics

Differences between the samples were analysed by ANOVA; a post hoc analysis was performed using the Tukey test (with a Bonferroni adjustment, in which p-values were multiplied by the number of tests being carried out). A related-samples Friedman’s two-way analysis of variance by ranks was used to compare the wPCDAI values for each patient at three time points and to compare the Shannon–Wiener diversity index at nine time points. A related-samples Wilcoxon signed rank test was used to compare the BW, BH and BMI z-scores at diagnosis and at the end of follow-up for each patient.

The relative abundance of operational taxonomic units (OTUs) was used to calculate the Shannon–Wiener diversity index with an aim to compare diversity between different sample groups. BioNumerics software (Applied Maths, Sint-Martens-Latem, Belgium) was used to perform cluster analyses based on the HhaI or MspI T-RFLP patterns. Due to a high number of variables in the OTU comparisons, p values were adjusted for multiple comparisons (p value was decreased by 5 times). Values below 0.01 were considered significant.

A permutational multivariate analysis of variance (PERMANOVA) (fixed effect, using the type III sum of squares and the unrestricted permutation of data with 999 permutations) was used to analyse the difference in abundance of OTUs between patients and sample times. Data were transformed (square root), and the Bray–Curtis measure was used to assess dissimilarities. Centroids were determined for every sample time point (baseline, on day two, on the last day of EEN, and every two months post-EEN for one year). A principal coordinates ordination analysis (PCO) was used to visually present the dissimilarities between patients and sampling time (nine sample times).

Statistical analysis was performed using SPSS 23.0 (Chicago, IL, USA) and Primer 7 software (Auckland, New Zealand).

The study was approved by the Ethics Committee of the Children’s Hospital Zagreb (IRB number: 21102014).