Setting

The follow-up study was conducted in Dhanusha district, in the lowland Terai region of Nepal. Dhanusha has around 768 000 inhabitants; most households are rural and over half are involved in agriculture.29 Maternal and child undernutrition are common in the Terai : 23% of women aged 15–49 have a Body Mass Index (BMI) below 18.5 kg/m², almost half (52%) of pregnant women are anaemic, and 36.7% of children under 5 are stunted.30 Opportunities for formal learning and stimulation during early childhood are limited: around half (49%) of children aged 36–49 months attend a preschool programme and 5% of children under 5 have access to children’s books.29 31

Design and participants

We followed up children born during the Randomised Controlled Trial (RCT) of antenatal micronutrient supplementation conducted in Janakpur, the district headquarters, between 2002 and 2004.27 The trial tested the effect supplement, the UNIMMAP, taken daily from the 12th week of gestation—at minimum—until delivery, compared with a daily supplement of iron (60 mg) and folic acid (400 µg) recommended by the government. The UNIMMAP contains vitamin A 800 µg, vitamin E 10 mg, vitamin D₅ µg, vitamin B₁ 1.4 mg, vitamin B₂ 1.4 mg, niacin 18 mg, vitamin B₆ 1.9 mg, vitamin B₁₂ 2.6 µg, folic acid 400 µg, vitamin C 70 mg, iron 30 mg, zinc 15 mg, copper 2 mg, selenium 65 µg and iodine 150 µg. The trial’s primary outcomes were birth weight and gestational duration. A total of 1200 pregnant women visiting the antenatal clinic of Janakpur Zonal hospital were randomly allocated to the intervention or control arm. Supplements in both arms looked, smelled and tasted identical. Women enrolled in the trial were followed up every 2 weeks through home and clinic visits, during which the study team assessed their adherence to supplements. The exclusion criteria were gestation beyond 20 weeks, multiple pregnancies, fetal abnormality detected by ultrasound and existing maternal illness that might compromise the pregnancy outcome. The trial team recorded 1069 deliveries and infants were followed up at birth and after 1 month. Women, their family members and the research team were masked to allocation.

Its primary outcomes were birth weight and gestational duration. A total of 1200 pregnant women who visited an antenatal clinic in Janakpur Zonal hospital were randomly allocated to the intervention or control arm, and followed up every 2  weeks through home and clinic visits. The trial team recorded 1069 deliveries, and infants were followed up at birth and after 1  month. Women, their family members and the research team were masked to allocation. Antenatal exposure to the UNIMMAP led to an increase of 77 g (95%  CI 24 to 130) in mean birth weight compared with IFA.27 32 33 Gains in weight were still present at 2.5 years, but disappeared by 8.5 years. 27 32 33

Procedures

We sought to locate children born during the parent trial and identified during the most recent follow-up in 2012.27 33 A member of the study team visited each family to obtain informed consent, interviewed mother and child, and invited them for a cognitive assessment. We made at least three attempts to invite and assess each child. We excluded children who were ill on all three attempts or had severe hearing, visual or motor impairments. To find and assess children who had migrated, we visited the nearby districts of Sarlahi and Siraha, as well as Kathmandu and Makwanpur district, where some children had been found previously.

Outcome measures

Our main outcome of interest was the Full Scale Intelligence Quotient (FSIQ) measured using the Universal Non-verbal Intelligence Test (UNIT).34 The UNIT was developed to test children irrespective of country of origin, sex or language. It includes subtests of symbolic memory, cube design, spatial memory, analogical reasoning, object memory and the ability to work through mazes. It is suitable for children aged 5–17 years, relies on non-verbal (gestural) instructions and feedback, takes around 45 min to administer and has been used previously in Nepal.35 The UNIT allows calculation of the FSIQ, a Memory Quotient (MQ), a Reasoning Quotient (RQ), which captures the ability to solve problems using information, a Symbolic Quotient (SQ), which assesses the ability to understand and process language, and a Non-symbolic Quotient (NSQ), reflecting the ability to perceive, recognise, sequence, organise and integrate information.34

We used a counting Stroop test to assess children’s executive function.36 The number–quantity Stroop task involved three conditions.37 38 In each condition, the children were asked to name the quantity of items in rows of one to seven identical items. In the baseline condition, children were asked to name the quantity of X in a row (eg, the correct response to XX is ‘two’). Second, in the congruent condition, they were asked to name the quantity of digits in a number with the equivalent digits (eg, 55555=‘five’). Finally, in the third, incongruent condition, they were asked to name the quantity of digits in a number with the non-equivalent digits (eg, 1111=‘four’). We recorded the time that children took to correctly name all items for each condition. We then derived two scores from these totals. An interference score was calculated by subtracting the baseline from the incongruent condition. This represented the potentially disruptive effect of automatic reading of the incongruous digits on quantity naming. A facilitation score was calculated by subtracting results for the congruent condition from those for the baseline, representing the enhancing effect of congruent digits.

Training of the testers was comparable to that for the US standardisation of the UNIT. The principal investigator (SD) was trained by two experienced UK-based psychologists (one clinical, one experimental, AK and FL) in the general principles of psychometric assessment and procedures specific to the UNIT and the counting Stroop test. She recruited two testers from Dhanusha district who were fluent in the local language (Maithili) and had a background in education and social science. The testers were trained for 15 days to administer the tests and practised under supervision for a week with children aged 10–13. We video-recorded the testers’ practice administrations of the UNIT. These were checked by the two experienced psychologists to ensure adherence to standardised testing procedures.

SD and the two other testers administered the UNIT and Stroop tests in an office with minimal distractions located in Janakpur. The testers recorded the children’s UNIT raw and scaled scores on a form with a unique child identification number. We checked each form individually and confirmed the scores using the proprietary CompuScore (V.1.1) UNIT software. These audited raw, scaled and standardised scores were double-entered into a Microsoft Access database. We identified discrepancies between first and second entries and resolved them by consulting the original forms.

To check the quality of UNIT assessments, we video-recorded 16 randomly selected assessments. Review of the videos by two experienced psychologists indicated that the cognitive testing procedures adhered closely to the standardised administration during data collection.

In addition to intelligence and executive function, we collected data on factors that could confound the effects of MMN supplementation on IQ, including the quality of the home environment, the child’s and parents’ educational attainment, the mother’s mental health and the child’s current nutritional and health status (height, weight, dietary recall, morbidity recall and mental health). We used a Shorr board, accurate to 1 mm, to measure children’s height and a Tanita HS302 solar scale, accurate to 100 g, to measure weight. We assessed mothers’ mental health with the 12-item General Health Questionnaire, which had been used previously in this setting.39 We used the Depression Self-Rating Scale (DSRS) and Screen for Child Related Anxiety Disorder (SCARED) tools to assess children’s mental health. The DSRS has been adapted and validated for use in Nepal.40 Finally, we used the Early Adolescent Home Observation Tool (HOME) to assess the quality of the children’s home environment. We followed the Home Inventory Administration Manual’s observation techniques and structured interview format.41 We removed one out of the 60 items in HOME because it was culturally sensitive: item 36 asked whether a parent had provided guidance to the adolescent about responsible sexuality and physical hygiene in the past year. In Dhanusha, parents do not commonly talk about sexuality with their children, sowe removed the question to avoid discomfort during the interview process. The interview format was translated into Nepali and Maithili, pretested and adapted before data collection. Questionnaire information was collected with smartphones using Open Data Kit.42 Interviewers received a month of training to ensure that they had the same understanding of tools and data collection processes. We piloted the tools in the community. Data collection was supervised and monitored throughout the follow-up period. Data were downloaded weekly and checked by SD. The data collection team and analyst (SD) were blind to the allocation of children in the parent trial. Figure 1 is a Mithila painting depicting the study team requesting consent from parents, interviewing a mother and child, and a child participating in the Stroop, UNIT and anthropometry.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Mithila painting depicting consent taking, interviewing, and a child participating in the Universal Non-verbal Intelligence Test, Stroop test and anthropometry.

Statistical methods

As the sample size was fixed by that of the parent trial, we estimated the smallest differences in mean UNIT scores that our study could detect with 80%–90% power, and an alpha of 0.05. Even with 16% attrition from the last follow-up (from 841 to 700 children), the study would have 87.1% power to detect a small (3.5) difference in mean IQ scores between groups assuming a SD of 15, or an effect size of 0.23 using Cohen’s d.

Primary analyses followed intention-to-treat principles. We compared the characteristics of participants identified and those lost to follow-up by trial allocation using descriptive statistics. We used WHO Child Growth Standards for children to generate Z scores for weight, height and BMI-for-age.43 We used independent two-sample t-tests to examine differences between intervention and control groups in means of standardised scores for the MQ, RQ, SQ, NSQ and FSIQ. We used linear regression models with and without adjustment for covariates. We examined the associations between each of the covariates in tables 2 and 3 with FSIQ scores and with one another through univariable analyses. We then selected covariates that were associated with FSIQ and were not collinear by visually examining their associations in a correlation matrix. A priori, and building on previous studies, we decided to carry out subgroup analyses to assess the effect of supplements on children’s FSIQ by adherence, maternal BMI, haemoglobin level in pregnancy and birth weight. Because the parent trial had detected a small differential effect of MMN on birth weight by sex, we also examined the association between exposure to the intervention and FSIQ by sex. Statistical analyses were conducted using STATA (V.13.1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Study profile. UNIT, Universal Non-verbal Intelligence Test.