The study was conducted from January 2010 to February 2012 in the low-to-middle socioeconomic neighborhoods of Tigri and Dakshinpuri in New Delhi, India. The total population was around 300,000; details of the population have been described previously [
12]. The current analysis has been done within the framework of a randomized double-blind placebo-controlled trial (NCT00717730 at
https://clinicaltrials.gov/ct2/show/NCT00717730) involving supplementation with folic acid and/or vitamin B12 or placebo for six months to 1000 children who were 6 to 30 months old at enrollment [
12]. Vitamin-D status was available for 960 children at baseline and in 243 at 6 months follow-up. For baseline, our analysis for neurodevelopment is restricted to the 401 children in whom ASQ-3 [
13] was administered, for physical growth to 960 children in whom anthropometry data were available at baseline and to 919 children in whom anthropometry data were available 6 months later. Of the 243 children, neurodevelopment data were available for 92 children and anthropometry data were available for all.
Assessment
Neurodevelopment was assessed 6 months after enrollment using the ASQ-3 which is a developmental screening tool constructed in the USA. [
13] The ASQ-3 consists of age-appropriate questionnaires, all containing 30 items divided into five subscales: Communication, Gross motor, Fine motor, Problem-solving and Personal-social, summing up to five subscale scores (range 0 to 60) and a total score (range 0 to 300). The construct and convergent validity of the translated ASQ-3 forms for the current setting were excellent, and our multiple models were able to explain more than 30% of the variability of the ASQ-3 scores. [
14]
Details of the Hindi translation and process of validation of ASQ-3, training and standardization methods have been described previously. [
15] Three trained field supervisors administered the ASQ-3 directly to the child at the study clinic in the presence of caregivers. The examiners elicited the relevant skills from the child during sessions using standardized materials. The caregiver served as an important contributor in supporting the child, eliciting behaviors and gave relevant information of the child’s development when necessary. During the 11 days of training, the field supervisors were standardized in performing the procedure, and they reached a high inter-observer agreement both during training and in the 10% quality control checks throughout the study. To assess the caregiver’s promotion of child development two questions were selected from the standardized assessment tool Home Observation for Measurement of the Environment (HOME) that were asked the caregivers during the session. [
16] One question was on “Mother’s belief that child’s behavior can be modified” and one was on “Mother’s encouragement of developmental advances”.
Trained field supervisors measured weight and length at baseline and after six months of supplementation. Weight was measured to the nearest 50 g using electronic scale (Digitron scale). Length was measured using locally manufactured infantometers reading to the nearest 0.1 cm.
Analytical procedures
Blood samples were obtained at baseline from all children; 3 mL blood was collected in an evacuated tube containing EDTA (Becton Dickinson). The plasma was centrifuged at ~450×g at room temperature for 10 min, separated, and transferred into storage vials and stored at −20
0 C until analyzed. Plasma concentration of vitamin-D was measured by quantitative electro-chemiluminescence binding assay, with detection of 25 OHD, the hydroxylated forms of vitamin-D2 (Roche Diagnostics, Mannheim, Germany) [
17] at the Department of Biochemistry, Christian Medical College, Vellore, India.
Statistical analysis
Proportions, means (SD) or medians (IQR) were calculated for categorical and continuous variables by vitamin-D status at baseline. Though The Institute of medicine concluded that for maximum bone health a blood level should be at least 20 ng/mL and the Endocrine Society’s Practice Guidelines recommended for maximum bone health a level should be above 30 ng/mL, we considered vitamin-D deficiency was defined at <10 ng/mL (25 nmol/L). [
18] We also ran a sensitivity analysis classifying baseline vitamin-D status as <10, 11–20, 21–29 and > = 30 ng/mL.
We used multiple regression and a “purposeful selection of covariates method” to identify variables that were associated with vitamin-D deficiency and our predefined outcomes. [
19,
20] These variables were used as adjustment variables in the multiple models where vitamin-D deficiency was the exposure variable. We also examined whether the predefined associations were modified by other variables using interaction terms (on a multiplicative scale) in the multiple regression models.
Multiple linear and logistic regression analyses were used to compare the total ASQ-3 and subscale-scores between the vitamin-D deficient and the vitamin-D non-deficient groups at baseline and at 6 months followup. In logistic regression models the total and subscale ASQ-3 scores were categorized at the 25th percentile. In these models, we adjusted for age of child, mother’s years of schooling, father’s years of schooling, log transformed annual family income, family structure, number of toys in the family, whether or not the family owns books, number of children in the family, hours of play with other children during the week, mother’s belief that child’s behavior can be modified, mother’s encouragement of developmental advances, weight-for-height Z score, weight for-age Z scores and intervention group.
We used multiple linear and logistic regression analyses to measure the association between vitamin-D deficiency and childhood physical growth at baseline and at 6 months follow-up. In logistic regression models, physical growth was categorized as wasting (< −2 Z scores weight-for-height/length), stunting (<−2 z score height/length-for-age) and underweight (< −2 Z scores weight-for-age). In the linear regression models, we used the Z scores of weight-for-height/length (WHZ), height/length-for-age (HAZ), weight-for-age (WAZ) as dependent variables. In these models, we adjusted for age, sex, breastfeeding status, family structure, log transformed annual family income, mother’s years of schooling, father’s year of schooling, baseline level of vitamin B12, folate and anemia status for baseline physical growth as well as, intervention group (placebo, folic acid, vitamin B12, or both) for 6 months later physical growth.
Statistical analyses were performed using STATA version 14 (Stata Corporation, College Station, TX).
We used generalized additive models in the statistical software R version 3.1.2 (The R Foundation for Statistical Computing, Vienna, Austria) to explore nonlinear associations between the vitamin-D status at baseline and HAZ score at baseline after adjustment for potential confounders [
21]. We also used generalized additive models to explore nonlinear associations between vitamin-D status at baseline and total ASQ-3 score after 6 months of follow - up.