Between October, 2001, and November, 2003, we did a community-based, placebo-controlled, randomised, 2• 2 factorial trial in the southern plains of Nepal (Nepal Nutrition Intervention Project, Sarlahi-4 [NNIPS-4]). The study population consisted of children 1−36 months of age who lived in households in the NNIPS catchment area of 30 Village Development Committees (VDCs; formerly called panchayats) in Sarlahi District along the border with Bihar State in northern India. This area is part of the flood plain of the Ganges river and its tributaries that drain from the Himalayas to the Bay of Bengal. It is typical of much of northern India and large parts of western Bangladesh and northern Pakistan. The participating VDCs are similar to others in the district with a predominance of traditional, rural, Hindu culture. The population is composed mostly of peasant farmers or labourers and their families and is considered a poor area even in Nepal; three-quarters of the population lives below the poverty line established by the government of Nepal.26
Every VDC is divided administratively into nine wards by the government. We have further divided these 270 wards into 426 sectors, to enable one local female community worker to visit all of the houses in her sector within a 20-h week. The number of sectors in each ward varies with the total population of the ward and the geographical layout. All of the sectors, and the villages within the sectors, have been mapped and the houses numbered. All children aged 1−35 months and living in households in the study area during the baseline enrolment round were eligible. In addition, all children born into households in the study area were eligible once 1 month old if that house was their primary residence.
We obtained oral informed consent on a community basis during meetings with community leaders. We also obtained verbal informed consent at the household level from the parents of eligible children. Occasionally, parents who initially refused to allow their children to be enrolled changed their minds during the follow-up phase of the study. We then enrolled these children. The Nepal Health Research Council, the Committee on Human Research of the Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA, and the Institutional Review Board of Cornell University, Ithaca, NY, USA, provided ethical approval.
We randomised children by sector, stratified by geographic area and in blocks of four, to receive one tablet daily (or half a tablet if • 1 year old), containing: iron (12·5 mg) and folic acid (50 • g), iron and folic acid plus zinc, or placebo. The sweet, vanilla flavoured supplements were specially formulated and manufactured as dispersible tablets by Nutriset (Malaunay, France) in conjunction with the Department of Child and Adolescent Health and Development at WHO, and were packaged in blister packs of seven tablets. The foil backing on the blister pack was imprinted with the treatment code. All children older than 6 months also received vitamin A as part of a national programme or, if missed, by study staff: those aged 12 months or older were given 200 000 IU of vitamin A every 6 months and those aged 6−12 months were given 100 000 IU.
We obtained data at the community, household, and individual level. Community-level data on the presence of economic, education, and health facilities were obtained in interviews with community leaders done just before the start of enrolment. Household-level data on socioeconomic status, health indicators, household structure, and housing material were obtained during the full population census also done before enrolment. All households that entered the trial for the first time because of a recent birth had their household interview at the time of child enrolment. On an individual level, study staff visited the children twice a week. They gave them a tablet directly and left enough tablets with the child's mother or caregiver to cover the daily doses until the next visit. Older children usually ate the tablet directly and mothers were instructed to dissolve the tablet in clean water or breastmilk for younger children. Study staff assessed compliance every week during their biweekly dosing and vital status assessment visits. The total number of tablets consumed by the child in the preceding week was recorded for every child. Investigators, study staff, and participants were unaware of assigned treatments. The Department of Child and Adolescent Health and Development at WHO, Geneva, Switzerland, kept the treatment assignment codes.
To confirm that the supplements used were active, we selected and tested the iron and zinc status of a sample of children aged 24 months or older after 12 months of follow-up. A blood sample was taken from participating children and the haemoglobin concentration measured with the HemoCue haemoglobinometer (HemoCue, Angelhom, Sweden). The remaining blood was centrifuged and the serum separated, collected, stored at • 10°C in Nepal, and then shipped in liquid nitrogen to Baltimore, MD, USA, where it was stored at • 70°C until analysis was undertaken. Serum ferritin was assessed with an ELISA assay (DELFIA system by Wallac, Gaithersburg, MD, USA).
Our primary outcome was all-cause mortality. The community workers recorded any deaths during their weekly visits to the participating families. The causes of death were determined by two physicians by independent review of information obtained by verbal autopsy with members of the immediate family after an appropriate period of mourning. If there was disagreement about the cause of death, a consensus meeting was held. Secondary outcomes included cause-specific mortality and the incidence and severity of diarrhoea, dysentery, and acute respiratory illness in two subsamples of 1200 children who were enrolled in the main trial and were younger than 24 months of age; we followed-up the first sample for 12 months and the second from enrolment in February, 2003, until the iron and folic acid-containing groups of the study were stopped in mid-November, 2003. Children in the morbidity subsamples were visited weekly. Mothers were asked about the onset and length of specific signs and symptoms for every day of the preceding week. The morbidities assessed included cough, fever, difficult or rapid breathing, diarrhoea, and dysentery. Visits to health-care providers in the previous week were also recorded. We diagnosed diarrhoea in individuals who produced four or more loose watery stools per day for one or more consecutive days, and persistent diarrhoea in those with watery stools for longer than 14 days. We defined dysentery as diarrhoea with blood or mucus in the stool on at least 1 day. We separated episodes of diarrhoea and dysentery by at least 3 symptom-free days. We diagnosed acute respiratory infection as one or more consecutive days of fever, cough, and difficulty breathing (all three symptoms had to be present on at least 1 day during the episode) with a minimum of 7 days between episodes.
To detect a reduction of 20% in overall mortality in the active treatment group with 80% power, a 2-sided type I error of 5%, a design effect of 1·23, accounting for 10% loss to follow-up, and based on the assumption of no interaction of iron, folic acid, and zinc on mortality, we calculated that approximately 24 000 person-years in every row (column) of the marginal comparisons—ie, iron and folic acid versus placebo or zinc versus placebo—were required. However, at the second meeting of the data and safety monitoring board about 14 months into recruitment, we noted that the mortality rate of children was much lower than originally expected. We therefore recalculated the sample size estimates. The final number needed was calculated at about 33 000 person-years in every row (column) of the marginal comparisons.
Analyses were by intention to treat. Children who migrated out of the study area or who refused further participation were censored at the time they left the study. We used SAS (version 8) and STATA (version 8.0) for statistical analyses. We compared treatment groups by baseline household, maternal, and child characteristics to assess imbalance after randomisation. We adjusted estimates of standard error to account for the clustered randomisation, using the generalised estimating equations approach.27
We used two approaches to assess the effect of treatment on mortality: the first estimated the incidence density of mortality with person-time as the denominator of the observed rates; and the second used survival analysis techniques such as Kaplan-Meier survival curves. We used Cox proportional hazard models to adjust for potentially confounding factors imbalanced in the treatment groups and to model potential effect modification. We estimated standard errors of the relative risks from proportional hazards models, using robust variance estimation to account for the clustered randomisation.28
We calculated incidence density rates of diarrhoea, dysentery, and acute respiratory infections with days during episodes excluded from the denominator of person-years at risk.
An independent data and safety monitoring board that included a paediatrician, a biostatistician, and the Member Secretary of the Nepal Health Research Council was formed and met four times to review the protocol and the data for safety and efficacy. There was no pre-established statistical stopping rule adopted by the data and safety monitoring board. However, detailed analysis was done only if the statistical evidence for treatment differences in mortality reached a p value of 0·2 or lower. At the fourth meeting in July, 2003, the board recommended that the iron and folic acid-containing groups of the trial be stopped because there was no evidence of a beneficial effect and the statistical power to detect significant differences in mortality between the treatment groups would be small by the time study recruitment and follow-up were completed. They concluded that further recruitment efforts would be better expended on increasing the zinc versus placebo groups of the trial. These groups will continue enrolment and follow-up in 2006. The ethical committees approved this decision and the protocol change was made in mid-November 2003.
So that the investigators did not become aware of treatment allocation, we gave a data file to an independent systems analyst who replaced the individual identifiers with a new, random set of identification numbers, filed the linked old and new identification information in a secure location, and then replaced the treatment codes with the actual treatment received. Here, we report the primary and secondary outcomes and their differences between the groups who were taking placebo, iron and folic acid, and iron and folic acid with zinc. The investigators, staff, and participants remain masked to the actual treatment codes.
This study is registered at clinicaltrials.gov, number NCT00109551.
Role of the funding source
The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.