Study design, site and participants

The design, site and participants of the iLiNS DYAD-Malawi trial have been described previously^{(Reference Ashorn, Alho and Ashorn14)}. In brief, this was a partially double-blind, individually randomised, controlled trial with three equal-size intervention groups conducted in the Mangochi District, south of Malawi. This area is largely rural, with the main source of livelihood being subsistence farming and fishing. The diet of people in the area consists predominantly of cereals mainly maize, with relatively small amounts of roots, tubers, fish, fruit and legumes^{(Reference Ndekha, Kulmala and Vaahtera21)}.

The trial participants were pregnant women presenting at antenatal clinics at four health facilities in the area between February 2011 and August 2012, including the district hospital in Mangochi, a semi-private hospital at Malindi and two public health centres at Lungwena and Namwera. Women were eligible if all of the following applied: no more than 20 completed weeks of pregnancy confirmed by an ultrasound scan; residence in the catchment area served by the four health facilities; no intention to travel out of the study area during the period of the intervention and signed or thumb-printed informed consent. Women were excluded if any of the following was present: (1) < 15 years of age; (2) requiring medical attention due to a chronic health condition; (3) known to be asthmatic or allergic to peanuts or to any substance; (4) severe illness warranting hospital referral or emergency medical care; (5) history of anaphylaxis; (6) pregnancy complications at the time of enrolment, including moderate to severe oedema, blood Hb concentration < 50 g/L, systolic blood pressure (BP) > 160 mmHg or diastolic BP > 100 mmHg; (7) history of being enroled in the iLiNS DYAD-Malawi trial during a previous pregnancy or (8) currently taking part in another clinical trial.

Intervention groups and study supplements

The intervention groups were (1) the iron–folic acid group (hereinafter, IFA) assigned to the standard nutritional supplementation for pregnant women in Malawi consisting of 60 mg/d Fe and 400 μg/d folic acid; (2) the MMN group assigned to 18 vitamins and minerals (including 20 mg/d Fe) and (3) the lipid-based nutrient supplements group (LNS) assigned to 20 g/d SQ-LNS containing similar micronutrients as the MMN group, and in addition, energy (494 kJ/d or 118 kcal/d), protein, essential fatty acids and the maximum amounts of Ca, Mg, P and K that could be included given technical and organoleptic constraints^{(Reference Arimond, Zeilani and Jungjohann11)}. The IFA and MMN groups served as controls.

The nutrient contents of the supplements used in the trial^{(Reference Ashorn, Alho and Ashorn14)} are presented in Table 1, and the rationale for the nutrient contents and concentrations of those for the MMN and LNS groups were reported previously^{(Reference Arimond, Zeilani and Jungjohann11)}. The supplements for the IFA and MMN groups were supplied by DSM Nutritional Products (Pty) Ltd in 10-capsule blister packs, while the SQ-LNS for the LNS group were produced by Nutriset S.A.S. (Malaunay, France) in individual 20 g sachets. Most of the vitamins and minerals in the MMN and SQ-LNS supplements were included at a dose of either 1x or 2x the RDA for pregnancy^{(Reference Arimond, Zeilani and Jungjohann11)}. We included 250 µg/d for iodine based on the WHO recommendation^{(Reference Andersson and de Benoist12)}.

Table 1

Composition of the supplements used in the iLiNS DYAD micronutrient supplementation trial in rural Malawi, by group

IFA, iron and folic acid: MMN, multiple micronutrient supplement; SQ-LNS, small-quantity lipid-based nutrient supplement. IFA was standard nutritional supplementation at the time of the study. Nutrient concentrations for SQ-LNS include contributions from the ingredients and from the multiple micronutrient premix. Total energy is equivalent to 118 kcal.

All supplements were intended to be consumed daily: for the IFA and MMN groups, 1 capsule was to be taken with water after a meal, and for the LNS group, 1 sachet of SQ-LNS was to be mixed with a small quantity of any food and consumed during the day. At the project office, supplements were stored at room temperature (20 °C–40 °C) in cardboard boxes away from light. In the homes, women were advised to store the supplement in dry places indoors.

Randomisation and enrolment

Pregnant women were considered enroled into the trial if they participated in a baseline assessment and were subsequently randomised into one of the three intervention groups. As previously described^{(Reference Ashorn, Alho and Ashorn14)}, the randomisation of the women was completed as follows: first, the Study Statistician at UC Davis created four randomisation code lists in blocks of nine, one list for each of the four health facilities serving as enrolment sites. To maintain blinding, each supplement or intervention group was coded using three different alphabetical letters between ‘A’ and ‘M’, so that field workers could identify women only by the alphabetical letters. Next, a researcher independent of the trial created randomisation slips for each enrolment site; each slip contained the letter code indicating the group assignment and a unique identification number for the participant. These slips were placed in opaque envelopes, which were sealed, labelled with the block numbers and individual enrolment numbers and stacked in increasing order of the block numbers. In order to not influence the women’s choices of envelopes when asked to make a pick from among a set of envelopes, we tried to hide the labels by writing in pencil on one side of the envelope that was never shown to the women.

At each enrolment, the randomiser shuffled the six topmost envelopes in the stack and asked the potential participant to choose one, which revealed the group assignment and identification number. The randomiser then returned the unused envelopes to the top of the stack. The process was repeated until all of the envelopes prepared for the enrolment site were used. When there were less than six envelopes left for a participant to pick from, the randomiser presented whatever number of envelopes that remained. At no point during randomisation, however, was it possible for the randomiser to guess the remaining allocations, since he or she had no knowledge of the randomisation scheme.

After randomisation, women received a 2-week supply of the assigned supplement along with the instructions on how to consume it. We told women not to consume more than the recommended dose each day even if they missed taking the supplement the previous day or days.

Data collection and follow-up procedures

During the baseline assessment, trained anthropometrists measured women’s weight (SECA 874), height (Harpenden stadiometer) and mid-upper arm circumference (Shorrtape, Weigh and Measure, LLC). Research nurses determined women’s malaria parasitaemia by rapid diagnostics tests (Clearview Malaria Combo), Hb concentration by Hemocue (HemoCue AB) and gestational age by fetal dimensions (biparietal diameter, femur length and abdominal circumference) and ultrasound examination (EDAN DUS 3 Digital Ultrasonic Diagnostic Imaging System, EDAN Instruments, Inc.). In addition, the research nurses obtained spot urine samples from women, which were frozen until analysed. Finally, study workers collected women’s background demographic and socio-economic information at enrolment.

Following enrolment, field workers visited women in the homes every 2 weeks, whereupon they delivered fresh supplies of supplements and monitored adherence by counting the numbers of delivered and recovered capsules or sachets. As reported previously^{(Reference Ashorn, Alho and Ashorn14)}, we suspended the distribution supplements to women in the LNS group for a period from 1 through 21 August 2012, on the advice of the trial’s DSMB due to a new recommendation involving the testing of LNS products for the presence of Cronobacter sakazakii. During this period, 160 pregnant women in the LNS group who missed their supplements for a number of days ranging from 1 to 20 were provided with Fe and folic acid capsules (as were women in the IFA group), in accordance with standard guidelines in Malawi. However, a total of 33 women who missed receiving their assigned supplement could not be reached during the temporary Fe and folic acid distribution, since they were not available at their homes during the time. We resumed the distribution of the intended supplements to the LNS group after 22 August 2012. At a laboratory visit at 36 weeks of gestation, study nurses collected spot urine samples as done previously.

Because of the apparent differences between the capsules (IFA and MMN) and the SQ-LNS, it was not possible to blind field workers and participants to women receiving these supplements. The study staff who collected or analysed the samples were, however, blinded to the group assignments.

Sample size calculation and data analysis

As with many continuous biochemical outcomes in the trial, the target sample size for women’s UIC at 36 weeks of pregnancy was based on detecting an effect size or Cohen’s d ^{(Reference Cohen24)} of ≥0·5 between any two groups, with a two-sided 5 % test and 80 % power. Thus, approximately 105 women per group (or 315 women for the 3 groups) were required, after taking into account up to 25 % attrition. To reduce the risk of missing data, the sub-sample for the UIC analysis was selected from among women for whom urine samples collected at baseline and at 36 weeks of gestation were both available. At 36 weeks of gestation, we had UIC values for 313 women in the three groups, which gave > 90 % power to detect an effect size (Cohen’s d) of 0·5 between any two groups.

Before we began the present analysis, we prepared and posted the statistical analysis plan at our website (www.ilins.org). In both the statistical analysis plan and the trial protocol described at ClinicalTrials.org, we listed women’s UIC as a secondary outcome intended to be analysed separately. We performed all analyses using SAS for Windows Release 9.4 and included women in the analysis as randomised, regardless of adherence to treatment.

We calculated the household assets index, housing index and Household Food Insecurity Access Scale score as proxy indicators of background socio-economic status, by using principal component analysis^{(Reference Coates, Swindale and Bilinsky25)}. We summarised the background characteristics at enrolment, by group, as mean ± sd or frequency (%). Because UIC are known to be not normally distributed⁽²⁶⁾, we summarised UIC at baseline and 36 weeks of gestation and the change in UIC from baseline to 36 weeks as median and first and third quartiles (Q1, Q3).

We assessed the impact of the intervention on women’s UIC at 36 weeks of gestation by comparing the three groups using ANCOVA (SAS PROC GLIMMIX) with Tukey adjustment for multiple comparison, after natural log transforming all UIC values^{(Reference Adu-Afarwuah, Young and Lartey20)}. We performed the ANCOVA twice^{(Reference Adu-Afarwuah, Young and Lartey20)}: in the first, we included log-transformed baseline UIC as the only pre-specified covariate in the model; in the second, we included additional pre-specified background covariates in the model, if they were significantly associated with the outcome (log-transformed UIC at 36 weeks of gestation) at ≤ 0·1 level of significance. Pre-specified background variables evaluated as potential covariates in bivariate models were gestational age at enrolment, years of formal education, parity (nulliparous or parous), household assets index, Household Food Insecurity Access Scale^{(Reference Coates, Swindale and Bilinsky25)} and season of enrolment (wet or not wet). From the ANCOVA models, we calculated the adjusted group geometric mean UIC at 36 weeks of gestation and their 95 % CI by back transformation. The geometric mean UIC is considered an approximate estimator of median UIC^{(Reference Thomas27,Reference Hauschke, Steinijans and Pigeot28)} . We examined whether there were significant interaction effects (P _{for interaction} < 0·1) between treatment and the pre-specified background variables on women’s UIC at 36 weeks of gestation.

To assess the adequacy of the women’s iodine intakes using WHO guidelines, we compared the group median UIC at baseline and the group median as well as geometric mean UIC at 36 weeks (approximately the median of the untransformed values) with WHO cut-offs, whereby median UIC <150, 150–249, 250–499 and ≥500 µg/L represent ‘inadequate’, ‘adequate’, ‘above requirements’ and ‘excessive – i.e. in excess of the amount required to prevent and control iodine deficiency’ iodine intakes, respectively⁽²⁶⁾. Median UIC from spot urine samples⁽²⁹⁾ and the cut-offs⁽²⁶⁾ are recommended for assessing iodine intakes in populations⁽²⁶⁾.

Statistics in the texts are median (Q1, Q3) or geometric mean (95 % CI). As previously reported^{(Reference Klevor, Adu-Afarwuah and Ashorn30)}, women’s adherence to supplement intake during pregnancy (i.e. median percentage of follow-up days women self-reported consuming the supplements) was 91·7 % for the IFA group; 91·0 % for the MMN group and 93·8 % for the LNS group^{(Reference Klevor, Adu-Afarwuah and Ashorn30)}.