The results from this study embedded within the ENID Trial are reported in accordance with STROBE guidelines.
Study population
The ENID trial (ISRCTN49285450) primarily examined whether early immune development can be improved through pre-natal and infant nutritional repletion. The trial followed pregnant women and their infants up to one year of age. The ENID-Growth study was an extension of the ENID trial that continued to follow the infants to two years of age. The ENID trial protocol has been described in detail elsewhere [17]. In brief, pregnant women (< 20 weeks gestation) from rural subsistence-farming villages located within The Gambia were recruited in early 2010, and randomised to one of four supplementation groups until delivery: 1) Iron-folate = standard care, 2) multiple micronutrients (MMN), 3) protein-energy (PE) + iron-folate, or 4) PE + MMN. Their infants were then randomised from 6 to 18 months of age to one of two supplementation groups: 1) lipid-based nutritional supplementation (LNS) + MMN, or 2) LNS only.
Infants in the main ENID trial were born between August 2010 and February 2014. For the current sub-study, infants born between May 2011 and December 2012, where plasma samples were available, were included (Fig. 1). All infants received the Expanded Programme on Immunisation as per Gambian government protocol.
Anthropometric measurements
Anthropometric variables collected at birth (within 72 h of delivery), and at clinic visits when the infants were aged 6, 9, 12, 18, and 24 months were used. Weight was measured to the nearest 0·01 kg using electronic scales and recumbent length was measured to the nearest 0·1 cm using a length board. Growth indicators including length-for-age z-score (LAZ), weight-for-age z-score (WAZ), and weight-for-length z-score (WLZ) were computed using WHO Anthro software (http://www.who.int/childgrowth/software/en/). Infants were characterised as stunted, wasted, or underweight if they had LAZ, WLZ, and WAZ scores, respectively, below − 2 SD from the median of the WHO reference population.
Aflatoxin exposure
Blood samples collected at infant ages 6, 12, and 18 months were used to measure aflatoxin-albumin adduct (AF-alb) concentrations at the University of Leeds. AF-alb concentrations in 250 μl plasma samples were measured using a competitive ELISA method [19]. The CV% had to be less than 25% between repeats. The assay’s limit of detection (LOD) was 3 pg/mg albumin. A value of 1·5 pg/mg albumin was assigned to samples with AF-alb concentrations below this limit.
IGF-axis proteins
Blood samples collected at 12 and 18 months were analysed for IGF-1 and IGF Binding Protein-3 (IGFBP-3) concentrations using IDS-iSYS IGF-1 and IGFBP-3 assays, with the IDS-iSYS Multi-Discipline Automated System (Immunodiagnostic Systems Holdings PLC, UK). The LOD levels for IGF-1and IGFBP-3 assays were 10 ng/mL and 80 ng/mL, respectively. The intra and inter assay CV% for the IGF-1 assay were 3·4 and 6%, and for the IGFBP-3 assay were 2·5% and 5·4%, respectively.
Infant feeding practice and morbidity
Field assistants visited the infants at home weekly, and administered a morbidity and feeding questionnaire to the primary caregiver (typically the mother). At this visit, the primary caregiver was asked if the infant had experienced any vomiting, diarrhoea, rapid breathing, fever, or cough in the past seven days. At these weekly visits, they were also asked to provide information on breastfeeding practices and the type and frequency of weaning foods the infant consumed.
Covariates
Household quality, used as an indicator of socio-economic status (SES), was assessed by a questionnaire that collected information on the material of the main structural components (floor, roof and walls) of the house of the mother. The questionnaire was completed in the participants’ homes and were conducted by trained field assistants. For each of the household structural components a list of materials was provided with a scoring guide 1 to 5, with 1 being the lowest score and 5 the highest. For example, for the floors of the house five different types of materials were listed, if earth/sand/mud was used for the floor of the house the lowest score of 1 was entered, if carpet was used the highest score of 5 was entered. A weighted score based on the household materials (multiplied by 0.2 for floor, 0.3 for roof and 0.5 for wall) was then computed for each infant. For analyses infants were then divided into tertiles and classified as belonging to a low, middle or high SES household (cut off at < 2.6 for low; 2.61–3.2 for medium; > 3.21for high).
Other potential confounders included: season of sampling (wet season, June–October, or dry season, November–May), age (months) when non-breast milk foods were introduced (i.e. cessation of exclusive breastfeeding), supplementation group infants were assigned to (LNS + MMN, or LNS only), and incidence of infant diarrhoea and infant morbidity (combined episodes of diarrhoea, vomiting, cough, rapid breathing and fever) in the first two years of life.
Statistical analysis
Statistical analyses were performed using SPSS version 22·0 (SPSS Inc., Chicago, IL) and STATA version 14 (StataCorp LP).
AF-alb was log transformed (lnAF-alb) and presented as geometric mean, GM (95% CI).
Relationship between aflatoxin exposure and infant growth
Three separate multilevel linear models (MLM) with maximum likelihood estimation were used to examine the relationship between the repeated measures (at 6, 12, and 18 months of age) of the three infant growth outcomes (LAZ, WLZ and WAZ) and lnAF-alb levels (time-varying covariate also measured at 6, 12, and 18 months). In each model lnAF-alb was modelled as a continuous variable. Measurement occasion was at level one, and individuals at level two. Random effects of the intercept and slopes were allowed, and an unstructured covariance matrix of the random effects was used. All adjusted models included the following covariates: season of sampling (measured at 6, 12 and 18 months), mother’s household quality, supplementation group, and age (months) of introduction of non-breast milk foods. With the assumption that AF-alb value at a given time point represents the average exposure in the previous 6 months (for example AF-alb value at month 12 represents the average exposure between month 6 and month 12) the above MLM models assesses the temporal relationship between infant growth and aflatoxin exposure. Mother’s education was not included as it was not very discriminatory.
Additionally, four separate multilevel linear spline models (MLSM) were used to examine the relationship between lnAF-alb and change in infant growth (WAZ, LAZ, WLZ, and height) at three time intervals (6 to 12 months, 12 to 18 months, and 18 to 24 months). These were added as spline models to allow the slopes to be estimated separately for different observation periods (as the infant growth was not linear during the observation periods). Spline models can also estimate the effect of aflatoxin on different age periods of infant growth. These models use lnAF-alb value at 6, 12 or 18 months as the baseline exposure level of the next 6 months to evaluate its effect on infant growth in the next 6 months. To increase the flexibility in modelling infant growth, a series of linear splines with knots were used to model change in infant growth at the different time intervals. In each model knot points were set at 12 and 18 months, which allowed different linear slopes from 6 to 12 months, 12 to 18 months, and 18 to 24 months, with these slopes varying between individuals. To determine the effect of aflatoxin exposure on change in infant growth for the three time periods, lnAF-alb (modelled as a continuous variable) at 6, 12, and 18 months and their interactions with the linear splines for the three time intervals were then included in the model. If lnAF-alb were related to change in infant growth within a period of time, the interaction would be significantly different from zero.
In the MLSMs, adjustments were made for potential confounders identified from previous studies, these included season of sampling (measured at 6, 12, 18, and 24 months), mother’s household quality, supplementation group, infant morbidity, and age (months) of introduction of non-breast milk foods. Random effects for the baseline body size and the change in infant growth between 6 and 12 months, as indicators of growth scores, measured at three time points (6, 9, and 12 months) within this interval, were included.
Relationship between aflatoxin exposure and IGF-axis proteins
The associations between lnAF-alb measured at 6, 12, and 18 months of age, and IGF-1 and IGFBP-3 measured at 12 and 18 months of age were examined using Pearson Correlation. Mixed ANOVAs were used to investigate whether change in IGF-1/IGFBP-3 from 12 to 18 months of age was associated with the interaction between the level of AF-alb measured at 12 months of age and time. AF-alb measured at 12 months of age was divided into ‘high exposure’ and ‘low exposure’ by means of a median split, and was the between-subject factor in each model. Time was the within-subject factor in each model, and represented IGF-1/IGFBP-3 measured at 12 and 18 months.