Settings and subjects
The present study was done as an independent sub-analysis of a large prospective, multinational longitudinal birth cohort study conducted across eight different nations across the world -‘The Etiology, Risk Factors and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development (MAL-ED) Network’[24]. The Indian study site was a densely populated urban slum in Vellore, South India [25]. For birth enrolment, the population of 12,000 was covered and pregnant women identified by a door-to-door survey. Consenting mothers were invited to participate in the study after an informed consent immediately after birth, by consecutive sampling. The exclusion criteria included multiple pregnancies, family’s existing plans to migrate out during the study period, medical comorbidities in the index child, another child already registered in the study, and mothers not available to provide necessary informed consent. Trained field workers visited recruited children twice a week till their second birthday for active disease monitoring and surveillance as per the MAL-ED protocol, with subsequent follow-ups. The initial birth cohort recruitment was conducted in Vellore between March 2010 and February 2012 and enrolled 251 newborns. The birth cohort recruited had similar characteristics to that of the study site [25]. The original birth cohort enrollment and consequent follow-ups were approved by the Institutional Review Board of Christian Medical College, Vellore and children were recruited at each stage after informed parental consent.
Measures
Bayley Scales of Infant and Toddler Development-III (BSID-III)
The Bayley Scales of Infant and Toddler Development-III (BSID-III) assesses development in the domains of motor, language, cognition and social skills between 1 and 42 months of age [26]. Prior to administration, the BSID-III was culturally adapted, translated, back-translated and piloted. The testing was conducted by a single trained clinical psychologist for each time point in a child-friendly site in the community clinic. Quality control was assured by reviews of video-recorded tests, with 10% being reviewed locally and 5-6% centrally [27]. BSID-III was administered at 6, 15, 24 and 36 months of age. For each time point, psychologist administered items pertaining to cognition, motor (gross motor and fine motor) and language (receptive and expressive) components. Socio-emotional scale was parent reported and correspondingly scored.
Cognition component measures exploration and manipulation of toys, concept formation, object relatedness and memory. Motor domain measures gross motor functions of sitting, walking, running, etc and fine motor functions of sensorimotor development and hand skills. Receptive language constitutes receptive abilities and comprehension of the child, while expressive language includes expressive abilities in spoken language and other communication modes. Socio-emotional domain analyses social and interactive skills of the child [26]. Individual domain quotient was calculated from the raw score using child’s chronological age.
The WAMI measure for socio-economic position
The WAMI is a simplified measure of socio-economic position (SEP) developed during the MAL-ED study, with components of access to improved water and sanitation, assets, maternal education and total household income [28]. The WAMI questionnaire was translated and piloted prior to the administration. A trained field worker visited the home and administered the measure at 6, 12, 18, 24 and 36 months of child’s age. The questionnaire for WAMI was translated to Tamil and back-translated before administration by a trained field worker. The final score was calculated from these variables, which was then converted to a standardized score ranging from 0 to 1.
The HOME Scale
The Home Observation for the Measurement of the Environment (HOME) scale (Infant/Toddler version) analyses the home environment of the child including stimulation and support and is considered the gold standard with good psychometrics [29, 30]. This observed measure analyses specifically mother-child interactions including responsiveness.
The modified version has six subscales consisting of total 48 items with subscales of appropriate play materials, avoidance of restriction and punishment, organization of the environment, parental involvement, responsiveness to parent, and variety in daily stimulation [31]. The measure was translated and piloted in the community prior to its administration. A trained social worker observed the home environment for 45-60 minutes in the morning and completed the HOME scale with supplementary information using a mother/caregiver interview as per the MAL-ED study protocol [32]. HOME scale was administered at 6, 24 and 36 months of child’s age.
Raven’s progressive matrices
The Raven’s progressive matrices, a scale of non-verbal reasoning ability was used to assess maternal cognition at 6-8 months of child’s age [33]. This culture-fair test was administered by a trained psychologist to mothers in a centrally located distraction-free room in the community clinic [27]. As per the MAL-ED protocol, a combination of Raven’s Standard Progressive Matrices and Coloured Progressive Matrices was used to analyse cognitive ability of observations, comparisons, analogy and spatial organization. Each item had a missing component in a pattern series of figures and mother was asked to identify the best fitting figure to fit into the pattern. Correct responses were scored as raw scores, which were used for analyses.
The Self Reporting Questionnaire-20
The Self Reporting Questionnaire-20 (SRQ-20), developed by the World Health Organisation to assess depressive symptoms in low-resource settings, was used to assess maternal psychological disturbances at 1,6, 15, 24 and 36 months of child age [34]. This measure has 20 items and each item can be answered as ‘yes’ or ‘no’. A trained social worker administered the measure at home and special care was taken to maintain privacy for the mother. The total score was calculated using 16 items, which yielded one factor structure [35].
Blood collection and analysis
Blood samples were collected at 7, 15, 24, 36 and 60 months. Samples were tested for hemoglobin (7, 15, 24, and 36 months); ferritin assays (at 7, 15 and 24 months) and blood lead levels (at 15, 24 and 36 months). Hemocue® Hb 301 kit (azidemethemoglobin method) was used to estimate hemoglobin using a drop of blood immediately after collection of blood in the community study clinic. The blood samples were instantly refrigerated using cold packs and transported to the research lab within two hours of blood collection. Serum ferritin, transferrin receptors and lead levels were tested using chemiluminescence immunoassay, immunoturbidimetry and Graphite Furnace Atomic Absorption Spectroscopy (GFAAS) method respectively in the Biochemistry department, Christian Medical College, Vellore. Being a national reference laboratory, standard calibration processes were followed, assuring quality control. As inflammation can influence serum ferritin levels, both transferrin (R) and ferritin (F) levels were utilized to find total body iron levels using the formula, where positive values indicated iron reserves [36]: see formula 1 in the supplementary files.
Data entry and analysis
Data entry was made in the double entry database system managed by Data Coordinating Centre (DCC) of the MAL-ED study. Data collection forms filled by the field workers were authenticated by the field supervisor before entering into the database [24]. Data analysis was done using Stata version 13 (StataCorp. 2013. Stata Statistical Software. Release 13. College station, TX: StataCorp LP).
Statistical analysis
The outcome variable, BSID scores, assessed at 6, 15, 24 and 36 months of age was summarized using mean and standard deviation (SD) under specific domains namely - cognitive, language, motor and socio-emotional domain. Independent predictors such as maternal depression score, maternal cognition and various domains of HOME inventory scale were summarized as mean and SD. Height-for-age and weight-for age scores below -2 SD were categorized as stunted and underweight respectively. Values of blood iron measured at 7, 15 and 24 months were considered for analysis and the missing values were replaced with the average of measurements available at the other time points. Blood lead levels measured at 15, 24 and 36 months were averaged to obtain mean lead level. Domain-wise score of cognitive development was compared across the time points using repeated measures ANOVA test. To measure the effect of independent predictors on cognitive development scores, we used regression analysis considering the following equation,
Cognition developmentij = αi + β1 Genderij + β2 Stuntedij + β3 Underweightij + β4 Mother’s cognition scoresij + β5 SEPij + β6 Body ironij + β7 Blood leadij + β8 Mother’s depression scoresij + β9-14 Domains of HOME Inventory scaleij + εij
where ‘i’ and ‘j’ refers to children and time points of measurements respectively, and, εij representing the random error. Since the independent variables measured at various time points tend to correlate at the individual level, we used generalized estimating equations – population averaged model for regression to account for clustering, instead of Ordinary least square (OLS) method. Further, GEE offered the scope for accounting for type of correlation in the variables by specifying the correlation structure in the analysis. For our analysis, we decided upon the type of correlation structure using Quasi Information Criteria (QIC) and the one with smallest QIC value was considered further [37]. The parameters specified in the final GEE model were - gaussian family, identity link function and exchangeable correlation structure. Beta co-efficients along with 95% confidence interval have been reported for the independent predictors. Model fit was assessed using Wald statistics and p less than 0.05 was considered as statistical significance.