This was an exploratory, randomized, double-blind, controlled study conducted between May 2013 and September 2015 in Thailand. The protocol and all accompanying material provided to the subjects, such as information sheets or description of the study used to obtain informed consent, were submitted to the following ethics review committees: Institutional Review Board, Chulalongkorn University; Institutional Ethics Review Committee, Royal Thai Army Medical Department; Human Research Ethics Committee, Thammasat University. Approval from the three Ethics Committees was obtained before the start of the study and was documented in a letter to the investigators specifying the date on which the committee met and granted the approval.
Written informed consent was obtained from all parents/caregivers before inclusion in the study. The study was registered in ClinicalTrials.gov (March 18, 2013; #NCT01813175). The study was conducted according to ICH-GCP principles, and in compliance with the principles of the ‘Declaration of Helsinki’ (59th WMA General Assembly, Seoul, October 2008) and with the Thai laws and regulations. Inclusion criteria were a gestational age between 37 and 42 weeks, infant age 43–65 days, and exclusive formula feeding for at least 1 week (except for the breastfed reference group). Exclusion criteria were, malnutrition, weaned before inclusion, malformations, use of systemic antibiotics or anti-mycotic drugs within 4 weeks prior to study entry, gastroenteritis or diarrhoea in the last 2 weeks prior to study entry. Sample size calculation methods and randomisation and unblinding procedures are reported in detail in the “Supplemental Information and Methods” section.
Eligible infants in the formula-fed group started a 2-week run-in period with regular non-hydrolysed cow’s milk based infant formula (Nutricia, The Netherlands). Infants, who had successfully completed the run-in period, were randomized to receive the control formula or either one of the two investigational formula; control formula supplemented with 0.8 g/100 ml scGOS/lcFOS and B. breve M-16V at a dose of either 1 × 104 cfu/ml (Syn4) or 1 × 106 cfu/ml (Syn6) for 6 weeks. After the intervention period, infants received control formula for a wash-out period of 2 weeks. Non-randomized, exclusively breastfed infants were included as a reference (Fig. 1).
Stool samples were collected at baseline (after run-in period and before start of the intervention), Week 6 (after intervention) and Week 8 (after wash-out). Stool samples were collected by the parents into stool containers provided by the investigators. Samples were frozen at temperature of − 15 to − 20 °C immediately after collection by the parents and kept at this temperature until transport to the hospital and storage at − 80 °C. Fluorescent in situ hybridization (FISH)28 was used to assess the relative abundance (or proportion) of seven major gut bacterial taxanomic groups (Total Bifidobacterium species, Bacteroides distasonis/Bacteroides fragilis, Eubacterium rectale/Clostridium coccoides, Lactobacillus/Enterococcus, Enterobacteriaceae, Atopobium, Clostridium histolyticum/Clostridium lituseburense. The proportion or ‘relative abundance’ of these targeted taxonomic groups was measured by comparison with the total abundance of bacteria. In short, fixated fecal samples were hybridized with the taxon specific probes and then analysed using an automated Olympus AX70 epifluorescence microscope equipped with image analysis software. The relative abundance (or proportion) of cells belonging to a specific bacterial taxon was determined at 25 randomly chosen positions on each well by counting all bacterial cells using a DAPI filter set and by counting the targeted bacterial taxon using a Cy3 filter set.
Targeted microbiota quantification by q-PCR29 analyses was used to assess the abundance of Bifidobacterium breve and Bifidobacterium breve M-16V and the potential pathogens Campylobacter jejuni, Clostridium difficile, Clostridium perfringens, Staphylococcus aureus, Enteroaggregative Escherichia coli (EAEC), Enteropathogenic Escherichia coli (EPEC).
Short chain fatty acid (SCFA) and lactate were measured by Gas Chromatography (GC). Safety parameters (anthropometry, gastrointestinal tolerance, serious and non-serious adverse events) were also investigated. A detailed study scheme is illustrated in Fig. 1. A detailed description of the methods, including the oligonucleotide sequences of the primers and probes used for FISH and q-PCR analyses, is available in the “Supplemental Information and Methods”.
Danone Nutricia Research will grant data access, to researchers that meet the criteria for access to confidential clinical study data and are compliant with the DNR Clinical Trial Dataset Sharing policy.
Analyses of continuous (and binary transformed) data were performed on the intention-to-treat (ITT) group. For safety data, the all-subjects-treated (AST) group was used. Continuous outcomes were modelled using a linear mixed-effect model for repeated measures (MMRM) including post-baseline and baseline measurements in the response vector, intervention, time and study site as fixed factors, intervention by time as interaction term and subject as a random effect. An unstructured covariance structure was used to model the correlation among repeated measurements. Supplemental Table S1 shows LS (Least Squares) estimates of differences in change from baseline between groups, Standard Error Estimates, 95% CI, and P-values for the linear mixed model key parameters measured at week 6 (Bifidobacterium, Eubacterium, pH, L-lactate, acetate, propionate, butyrate). Covariate assessment was performed for the analysis of Bifidobacterium in order to identify environmental factors (e.g. stool frequency, use of antibiotics and mode of delivery) that could potentially influence the estimate of treatment effect. The assessment was carried out by adding a single covariate into the linear mixed-effect model and evaluating the change in treatment effect estimate (10% or more change was considered relevant). In case the total number of non-detected measurements (or measurements below the limit of detection) for a specific parameter exceeded 30% of the data in at least one of the groups (for each comparison), the data were transformed into binary presence/absence (detected/non-detected) type of data. Prevalence of detected measurements was modelled instead, using a generalized linear mixed model (GLMM) with binomial distribution and a logit link function with study site as a fixed factor, intervention by time as interaction term, and subject as a random effect. Treatment comparisons were evaluated against control using a two-sided 95% confidence interval with corresponding p-value.
In addition to the univariate analyses performed for each (continuous or binary) parameter, Redundancy Analysis (RDA) constrained ordination was applied on the set of Hellinger-transformed fish data with the fish parameters as response variables and treatment as explanatory variable in order to assess the effect of treatment on the microbial assemblage composition. An ANOVA like permutation test30 was used to evaluate statistical significance of the treatment differences based on the resulting model. All analyses were performed using SAS (Enterprise Guide Version 4.3, SAS Institute, NC) except for RDA, which was performed using the ‘Vegan’ package in R (R software version 3.4.1, R Foundation for Statistical Computing, Vienna, Austria).