Infant gut microbiota characteristics generally do not modify effects of lipid-based nutrient supplementation on growth or inflammation: secondary analysis of a randomized controlled trial in Malawi

An unhealthy gut microbial community may act as a barrier to improvement in growth and health outcomes in response to nutritional interventions. The objective of this analysis was to determine whether the infant microbiota modified the effects of a randomized controlled trial of lipid-based nutrient supplements (LNS) in Malawi on growth and inflammation at 12 and 18 months, respectively. We characterized baseline microbiota composition of fecal samples at 6 months of age (n = 506, prior to infant supplementation, which extended to 18 months) using 16S rRNA gene sequencing of the V4 region. Features of the gut microbiota previously identified as being involved in fatty acid or micronutrient metabolism or in outcomes relating to growth and inflammation, especially in children, were investigated. Prior to correction for multiple hypothesis testing, the effects of LNS on growth appeared to be modified by Clostridium (p-for-interaction = 0.02), Ruminococcus (p-for-interaction = 0.007), and Firmicutes (p-for-interaction = 0.04) and effects on inflammation appeared to be modified by Faecalibacterium (p-for-interaction = 0.03) and Streptococcus (p-for-interaction = 0.004). However, after correction for multiple hypothesis testing these findings were not statistically significant, suggesting that the gut microbiota did not alter the effect of LNS on infant growth and inflammation in this cohort. A randomized, controlled, partially blinded, parallel-group clinical trial known as the International Lipid-based Nutrient Supplements DYAD (iLiNS-DYAD) trial was conducted in the Mangochi district of rural Malawi1,74. The main study hypothesis was that children whose mothers were provided with LNS during pregnancy and for 6 months after delivery and who themselves received LNS from 6 to 18 months of age would have a higher mean length at 18 months than children whose mothers received either IFA during pregnancy only or MMN supplementation during pregnancy and lactation and who themselves received no LNS. For primary outcome analysis and study information, please refer to Ashorn et al.74. The enrollment to the study took place in one public district hospital (Mangochi), one semiprivate hospital (Malindi), and 2 public health centers (Lungwena and Namwera) in Mangochi District, southern Malawi. The target population comprised pregnant women who came for antenatal care at any of the study clinics during the enrollment period and met the inclusion criteria75. Between February 2011 and August 2012, the iLiNS team members approached a total of 9,310 women, from whom 1,391 (14.9%) were enrolled in the trial and were randomly assigned to 1 of the 3 intervention groups. Of these, 869 women were assigned to the complete intervention and follow-up until 18 months after delivery. Singleton children born to these women formed the sample for the present study. Infants born to the remaining 522 women who were assigned to pregnancy intervention only were not included in the present analyses (Fig. 1). Details of the study, including the original sample size calculation, are described elsewhere75. At enrollment, study personnel collected data on socio-demographic status, maternal age, height, body mass index (BMI), parity, education, HIV status, hemoglobin concentration, household assets, food security, source of drinking water, access to sanitary facilities, and season. Household asset and food security indices were created as previously described76,77. Details of the trial are available at the National Institutes of Health (USA) clinical trial registry (www.clinicaltrials.gov), under the registration number {“type”:”clinical-trial”,”attrs”:{“text”:”NCT01239693″,”term_id”:”NCT01239693″}}NCT01239693 (11/10/2010). The trial was conducted in adherence with the Good Clinical Practice guidelines and ethical standards of the Helsinki Declaration. The trial protocol was approved and ethical clearance to conduct the study was granted by the University of Malawi College of Medicine Research and Ethics Committee (COMREC) and the ethics committee at Tampere University Hospital District, Finland. Informed consent was obtained from each participant before being enrolled into the study. An independent data safety and monitoring board monitored the incidence of suspected serious adverse events during the trial. Women were randomly assigned to three groups as described previously74: iron and folic acid during pregnancy only (IFA), a multiple micronutrient (MMN) tablet during pregnancy and the first 6 months postpartum, or LNS during pregnancy and the first 6 months postpartum. Briefly, an independent researcher (not involved with the trial) created individual randomization slips in blocks of 9. The slips were then packed in sealed, numbered, and opaque randomization envelopes stored in numerical order. Enrolled women were asked to choose 1 of the top 6 envelopes in the stack, and the contents of each chosen envelope indicated her participant number and group allocation. A statistician not involved in the study maintained the intervention code, which was not broken until all laboratory and statistical analyses of primary outcomes were performed. The IFA/placebo and MMN capsules were identical in appearance. Children born to mothers in the IFA and MMN groups received no supplementation; children in the LNS group received small quantity lipid-based nutrient supplementation (SQ-LNS) from 6 to 18 months. The LNS given to infants differed from that given to mothers as the nutrient content was designed to meet the needs of infants21. All stool sample collection and sequencing occurred prior to the current data analysis. Sample collection and sequencing was performed as previously described61. Briefly, infant stool samples were collected at 6 months, 12 months, and 18 months of age. Stool samples were collected in the home, the morning of the clinic visits and frozen at − 20 °C before being transported to the central clinic in Mangochi and stored at − 80 °C. After DNA extraction, the variable region 4 (V4) of bacterial 16S rRNA was amplified by PCR and sequenced using Illumina MiSeq. QIIME 1 was used to cluster reads into operational taxonomic units (OTUs) at 97% sequence identity using the May 2013 Greengenes database78. Taxonomy was assigned using the Ribosomal Database Project classifier 2.4. Raw counts were rarefied to 10,000 reads as determined by construction of a rarefaction curve and singleton OTUs were filtered out. Abundance counts were normalized using total sum scaling (TSS). All outcomes were assessed as continuous variables. This included LAZ, WAZ, WLZ, HCZ, CRP, and AGP. Age- and sex-standardized anthropometric indices (LAZ, WAZ, WLZ, and HCZ) were calculated using the WHO Child Growth Standards79. Growth outcomes were assessed as the change from 6 to 12 months of age while inflammatory outcomes were assessed as absolute values at 18 months of age. Potential effect modifiers chosen for the analysis included relative abundance of Bifidobacterium, Lactobacillus, Clostridium, Dorea, Enterococcus, Escherichia, Faecalibacterium, Ruminococcus, and Streptococcus as well as patterns of microbiota composition including E/B ratio, F/B ratio, α-diversity (Shannon index), richness (Chao1), and MAZ (Supplementary Table ​Table1).1). The E/B and F/B ratios were analyzed separately by their component parts (e.g. Enterobacteriaceae and Bacteroidaceae; Firmicutes and Bacteroidetes) since we observed abundances below the detectable limit of taxa which made the ratios incalculable for 33% and 12% of children, respectively. Dichotomous variables were created to categorize taxonomic features into values above and below the median for all taxa except Bifidobacterium, which displayed a normal distribution. This allowed us to uniformly handle the skew of microbiota measures. For taxa for which the median was zero, such as Dorea and Faecalibacterium, we categorized into presence or absence, or abundance below the detectable limit. Statistical analyses were performed in R (version 3.4.0)80. Maternal IFA and MMN intervention groups were combined into one control group because neither supplement was provided to the child and neither contained fats. Inflammation outcomes, AGP and CRP, were natural log transformed. Potential effect modifiers were assessed with an interaction term between the effect modifier variable and intervention group in covariate adjusted ANCOVA or logistic models. Potential adjustment covariates included site of enrollment, estimated pre-pregnancy maternal BMI, maternal height, maternal education, maternal age, maternal HIV status, delivery method, parity, household assets, food security index, and season at 6 months postpartum and were included if significantly associated with the outcome at 10% level of significance in a bivariate analysis. We controlled for multiple hypothesis testing using Benjamini–Hochberg corrections, using a 0.15 false discovery rate as the threshold for significance.