The Journal of Nutrition
Community and International Nutrition
Maternal and Infant Lipid-Based Nutritional
Supplementation Increases Height of
Ghanaian Children at 4–6 Years Only if the
Mother Was Not Overweight Before
Conception
Sika M Kumordzie,1 Seth Adu-Afarwuah,2 Mary Arimond,1,3 Rebecca R Young,1 Theodosia Adom,4
Rose Boatin,4 Maku E Ocansey,1 Harriet Okronipa,1 Elizabeth L Prado,1 Brietta M Oaks,1,5 and Kathryn
G Dewey1
1Program in International and Community Nutrition, Department of Nutrition, University of California, Davis, CA; 2Department of
Nutrition and Food Science, University of Ghana, Ghana; 3Intake – Center for Dietary Assessment, FHI 360, Washington, DC; 4Nutrition
Research Centre, Radiological and Medical Sciences Research Institute, Ghana Atomic Energy Commission, Legon, Ghana; and
5Department of Nutrition and Food Sciences, University of Rhode Island, Kingston, RI
ABSTRACT
Background: Few studies have evaluated the long-term effects of nutritional supplementation during the first 1000 d
of life. We previously reported that maternal and child lipid-based nutrient supplements (LNS) increased child length by
18 mo.
Objective: The aim of this study was to examine the effects of LNS on later growth and body composition at 4–6 y of
age.
Design: This was a follow-up of children in the International Lipid-based Nutrient Supplements (iLiNS)-DYAD trial in
Ghana. Women (n= 1320) at ≤20 weeks of gestation were randomly assigned to: 1) iron and folic acid during pregnancy
and 200 mg calcium/d for 6 mo postpartum, 2) multiple micronutrients (1–2 RDA of 18 vitamins and minerals) during
both periods, or 3) maternal LNS during both periods plus child LNS from 6 to 18 mo. At 4–6 y, we compared height,
height-for-age z score (HAZ), and % body fat (deuterium dilution method) between the LNS group and the 2 non-LNS
groups combined.
Results: Data were available for 961 children (76.5% of live births). There were no significant differences between
LNS compared with non-LNS groups in height [106.7 compared with 106.3 cm (mean difference, MD, 0.36; P = 0.226)],
HAZ [−0.49 compared with −0.57 (MD = 0.08; P = 0.226)], stunting (< -2 SD) [6.5 compared with 6.3% (OR = 1.00;
P = 0.993)], or % body fat [15.5 compared with 15.3% (MD = 0.16; P = 0.630)]. However, there was an interaction
with maternal prepregnancy BMI (kg/m2) (P-interaction = 0.046 before correction for multiple testing): among children
of women with BMI < 25 , LNS children were taller than non-LNS children (+1.1 cm, P = 0.017), whereas there was no
difference among children of women with BMI ≥ 25 (+0.1 cm; P = 0.874).
Conclusions: There was no overall effect of LNS on height at 4–6 y in this cohort, which had a low stunting rate, but
height was greater in the LNS group among children of nonoverweight/obese women. There was no adverse impact of
LNS on body composition. This trial was registered at clinicaltrials.gov as NCT00970866. J Nutr 2019;149:847–855.
Keywords: growth, body composition, lipid-based nutrient supplements, follow-up, prenatal supplementation
Introduction of child anthropometric indices to later health may depend on
the index examined. Stunting is a risk factor for diminished
Nutrition is essential for child growth and development survival, learning capacity, and productivity (2, 5). On the
throughout the first 1000 d of life (conception to the child’s other hand, rapid increases in weight and BMI during the
second birthday) and beyond (1–3). Early childhood nutrition preschool years are associated with adult obesity and altered
has been linked to several health outcomes in later life body composition in adolescence and adulthood (6).
including obesity and chronic diseases such as type 2 diabetes, Although several randomized trials have examined the
hypertension, and cardiovascular disease (4). The relationships effects of nutritional supplementation in early life, very few have
Copyright©C American Society for Nutrition 2019. All rights reserved. This is an Open Access article distributed under the terms of the Creative Commons Attribution
License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is
properly cited.
Manuscript received September 26, 2018. Initial review completed October 31, 2018. Revision accepted January 8, 2019.
First published online April 29, 2019; doi: https://doi.org/10.1093/jn/nxz005. 847
included follow-up of long-term consequences. Two landmark supplementation, (2) daily MMN tablets (1–2 RDA of 18 vitamins
trials of maternal plus child food supplementation conducted in and minerals) during pregnancy and the first 6 mo postpartum and no
the 1970s included assessments several years post intervention. infant supplementation, or (3) daily 20 g (118 kcal) lipid-based nutrient
In Guatemala (7), pregnant and lactating women and their supplements (LNS) during pregnancy and the first 6 mo postpartum
children from birth to 7 y of age received either a high- followed by infant 20 g/d LNS supplementation from 6 to 18 mo of age.
protein, high-energy supplement called “Atole”or a nonprotein, The maternal LNS had the same micronutrient content as the MMN,
plus calcium, magnesium, phosphorus, potassium, and macronutrients
low-energy supplement called “Fresco,” both fortified with (essential fatty acids and a small amount of protein), and the infant LNS
micronutrients. In adolescence, children in the Atole group were had the same macronutrients and 22 micronutrients based on infant
taller, heavier, and had higher fat-free mass (FFM) than those in Recommended Nutrient Intakes (11). Data collection in the main trial
the Fresco group. In Colombia, there was a sustained positive ended for women at 6 mo postpartum and for the children at 18 mo
effect of maternal and child food supplementation on child of age.
height and weight 3 y after the intervention period ended (8).
The International Lipid-based Nutrient Supplements Follow-up when children were 4–6 y of age
(iLiNS)-DYAD trial in Ghana is 1 of the first randomized trials After themain trial ended in 2014, an update of contact informationwas
since the 1970s to assess the impact of combined maternal and undertaken in preparation for a future follow-up study. The follow-up
infant supplementation on child growth and other outcomes (9, study occurred in 2016 when the children were 4–6 y of age.
10). The iLiNS-DYAD trial tested the efficacy of small-quantity
lipid-based nutrient supplements for both mothers and infants, Participants at follow-up.
compared with 2 other treatments for mothers only [prenatal For the follow-up at 4–6 y of child’s age, we sought to locate all children
iron and folic acid (IFA) followed by postnatal calcium (200 born to pregnant women who had been randomized to 1 of the 3
mg/d) supplementation, or multiple micronutrients (MMN) intervention groups of the iLiNS-DYAD trial, regardless of whether
both prenatally and postnatally], with regard to pregnancy their mothers were lost to follow-up before delivery or they themselves
outcomes and child growth to 18 mo of age. At birth, infants or their mothers were lost to follow-up after delivery and before the
born to women in the LNS group had significantly greater birth end of the main trial. Excluding misdiagnosed pregnancies (n = 5),
weight, weight-for-age, and BMI-for-age z scores compared to miscarriages, and stillbirths (n = 66), and children who died before the
end of the main trial (n = 27), 1222 children were potentially eligible to
children born to women in the IFA group but not the MMN participate. Details of participant enrollment and eligibility are shown
supplement group; among infants of primiparous women, in Figure 1.
the LNS group had greater weight and length compared to The follow-up study protocols were approved by the Institutional
both control groups (10). By 18 mo, there was a significant Review Boards of the University of California, Davis, the Ethics
positive effect of LNS on length, stunting prevalence, and Committee for the College of Basic and Applied Sciences at the
weight compared to the 2 control groups combined (9). Our University of Ghana, and the Ghana Health Service Ethical Review
objective herein is to examine the effect of maternal and Committee. Both the main trial and follow-up were registered at
infant supplementation with LNS on child growth and body clinicaltrials.gov as NCT00970866. Written informed consent was
composition at 4–6 y of age in the iLiNS-DYAD Ghana cohort. given by the primary caregiver before data collection.
Data collection procedures.
Before data collection for the follow-up began, we conducted a
Methods pilot study using the deuterium dilution method to determine the
Location and study design of the main trial equilibration time for a deuterium dose in 4–6-y-old children. The pilot
The iLiNS-DYAD Ghana trial was conducted in the Yilo and Lower study followed the procedures outlined by the International Atomic
Manya Krobo districts of the Eastern Region of Ghana between Energy Agency for determination of total body water (TBW) using a
December 2009 and March 2014. Details of the study are published Fourier transform infrared spectrometer (12). The sample size for the
elsewhere (10). Briefly, the iLiNS-DYAD study was a partially double- pilot study was 15 children. The children consumed a dose of deuterium
blind randomized controlled trial which enrolled 1320 women aged based on their body weight. No adverse effects of deuterium oxide at
≥18 y at ≤20 weeks of gestation attending antenatal clinics in 4 main this amount have been reported in humans (12). Saliva samples were
health facilities in the study area. The women were randomized to 1 of collected at baseline before the dose, 2 h after the dose, and every 30
3 treatments: (1) daily IFA tablets during pregnancy and a 200 mg/d min until 4 h after the dose was given. Data were analyzed by plotting
calcium tablet (placebo) during the first 6 mo postpartum and no infant a graph of mean deuterium concentration against time of post-dose
sample collection, which showed a plateau between 2.5 h and 3.5 h.
To determine the equilibration time, we calculated each child’s % CV
Funded by a grant to the University of California, Davis from the Bill & Melinda for deuterium enrichment for each pair of time points, between 2 h and
Gates Foundation. However, the findings and conclusions contained within are 2.5 h, 2.5 h and 3 h, 3 h and 3.5 h, and 3.5 h and 4 h. Equilibration is
those of the authors and do not necessarily reflect the positions or policies of the considered to have occurred when the % CV is ≤2% (12). The mean
Bill & Melinda Gates Foundation. Peter J Shields and Henry A Jastro Research % CV was 1.46, 0.61, 0.74, and 1.77 for the 2 h and 2.5 h, 2.5 h and 3
Award from the Graduate Group in Nutritional Biology, University of California,
Davis (Pilot study). h, 3 h and 3.5 h, and 3.5 h and 4 h time points, respectively. Based on
Author disclosures: SMK, SA-A, MA, RRY, TA, RB, MEO, HO, ELP, BMO, and this analysis, we chose 2.5 h and 3 h after dose administration as the
KGD, no conflicts of interest. equilibration times to be used in the follow-up study.
Supplemental Tables 1–4 and Supplemental Figure 1 are available from the For the follow-up, data collection was done at a central location.
“Supplementary data” link in the online posting of the article and from the same We used Open Data Kit software and programmed our questionnaires
link in the online table of contents at https://academic.oup.com/jn/. on tablets (13) with quality checks built into the system to minimize
Address correspondence to SMK (e-mail: smkumordzie@ucdavis.edu). entry of implausible values. A questionnaire was administered to update
Abbreviations used: AFA, arm fat area; AMA, arm muscle area; FFM, information provided at enrollment into the main trial. Anthropometric
fat-free mass; FM, fat mass; HAZ, height-for-age z score; IFA, iron and measurements were taken by 2 anthropometrists who had been trained
folic acid; iLiNS, International Lipid-Based Nutrient Supplements; INCAP,
The Institute of Nutrition of Central America and Panama; LNS, lipid-based to take similar measurements in the main trial, 1 of whom was
nutrient supplements; MMN, multiple micronutrients; MUAC, midupper arm considered the “gold standard” for standardization sessions. The
circumference; TBW, total body water; TSF, triceps skinfolds; WAZ, weight-for- anthropometrists were retrained and standardized at the beginning of
age z score. data collection and every quarter until the end of data collection (14).
848 Kumordzie et al.
>20 weeks gestation (197) 
1575 recruited 
Refusal by woman (24) 
1320 enrolled into main Loss of pregnancy (9) 
trial Misdiagnosed pregnancy (9) 
Acute illness (8) 
Randomly Assigned  Moved/planned to move (2) 
Other (6) 
IFA MMN LNS 
n = 441 n = 439 n = 440 
Misdiagnosed Misdiagnosed 
pregnancy pregnancy (n = 4) 
(n = 1 ) Miscarriage 
Miscarriage (n = 12 )
(n = 25)
Targeted for follow-up Targeted for 
n = 854 follow-up 
n = 424 
Still birth 
(n = 22 ) Still birth (n = 7) 
Child died Child died 
(n = 15) (n = 12) 
Eligible for re-enrollment at age 4-6y Eligible for  re-
n = 817 enrollment at age 4-6y n = 405 
Not traced (n = 65 ) Not traced (n = 31 )
Refused (n = 35 ) Refused (n = 8 )
Child died after Child died after 
main trial (n = 3 ) main trial (n = 2 )
Moved (n = 52 ) Moved (n = 12 )
Re-enrolled at age 4-6y Re-enrolled 
n = 662 at age 4-6y n = 352 
Lost to follow- Lost to follow-
up (n = 39) up (n = 14) 
Children measured Children 
n = 623 measured n = 338 
FIGURE 1 Study profile of the International Lipid-Based Nutrient Supplements (iLiNS)-DYAD Ghana trial. IFA, iron folic acid; LNS, lipid-based
nutrient supplement; MMN, multiple micronutrients.
Height was measured using a stadiometer (Seca 217) to 0.1 cm, triplicate if the first 2 measurements differed by a predefined amount:
weight to the nearest 50 g (Seca 875 scale),midupper arm circumference 0.1 kg for weight, 0.5 cm for height and MUAC, and 2 mm for TSF.
(MUAC) to the nearest 0.1 cm using a Shorr tape, and triceps skinfolds For TBW determination, children were given a dose of deuterium
(TSF) to 0.2 mm with Lange calipers according to WHO standard based on weight. Children <20 kg received a 6 g dose, whereas those
procedures (15). All measurements were taken in duplicate and in ≥20 kg were given a 10 g dose as recommended by the International
Early nutrition supplementation and later growth 849
Atomic Energy Agency (12). All doses were prepared in bulk as a 1 in 5 uncertainty we were unable to use sex-specific cutoffs for %FM, and
dilution. If the child had been sick (defined as fever, cold, cough,malaria, instead chose a somewhat arbitrary cutoff (20%) that was in between
hospitalization) in the previous 7 d before the day of data collection, the the 85th percentile reference values for boys and girls 5–6 y of age [85th
TBW assessment was rescheduled. Because the children were young, we percentile, 18.6% and 19.5% for 5-y-old and 6-y-old boys, and 21.5%
required that all children be given breakfast before they came for sample and 23.0% for 5-y-old and 6-y-old girls, respectively] (19) and coincided
collection, and the dose was administered at least 2 h after breakfast. with the 85th percentile in our sample.
A snack (150 g of a chocolate drink and a packet of biscuits) of The sample size for the primary outcome of height was based on
<300 kcal was also provided 1 h after the dose to standardize what detecting an effect size, Cohen’s d (20), of ≥0.25 in mean height. This
they were offered. Any leftovers were weighed and recorded. Saliva resulted in a minimum sample size of 198 per group to detect the
samples were stored in a −33◦C freezer and transported in ice chests to difference with 80% power and α = 0.05. (16) Our goal, however,
the laboratory at the Ghana Atomic Energy Commission for analysis. was to measure all of the children we could locate, which provided
Each sample was measured in duplicate and the mean deuterium sufficient power to test our second hypothesis regarding % body fat,
concentration for each time point was calculated. Each morning and and to evaluate other secondary outcomes within this longitudinal
at different times during the sample analysis, performance of the cohort.
equipment was checked by running quality control samples. A statistical analysis plan was posted on the iLiNS Project website
Percent CV was calculated using the means of the 2.5 h and 3 h (www.ilins.org) before data analysis. Analysis was carried out based on
time points. If the % CV was >5%, samples were rerun and if sample a complete case intention-to-treat principle using SAS version 9.4. All
volume was not sufficient for the rerun or the % CV remained >5%, tests except the noninferiority hypothesis were 2-sided at a 5% level of
only the enrichment at 3 h was used in the calculation of TBW. For significance.
subjects without the 3 h sample, enrichment at 2.5 h was used in the All continuous data were examined by univariate analysis to
calculation of TBW. identify outliers. The TSF, AMA, and AFA variables were not normally
To determine TBW, the average of the means at 2.5 and 3 h was distributed, therefore they were log-transformed.
taken and used together with the amount of dose administered and ANCOVA was used to analyze the data testing the null hypothesis
dilution space of the body in the calculation of TBW. Where there was of no difference in means for continuous variables. All models included
only a value for 1 time point, that value was used to calculate TBW. child age at follow-up. Fully adjusted models were also adjusted for
FFMwas then calculated from TBW and age- and sex-specific hydration a set of prespecified potential covariates. In addition to child age, the
factors (12). The difference between body weight and FFM gives fat potential covariates included: sex of child, maternal gestational age at
mass (FM). enrollment, nulliparity, maternal education and estimated prepregnancy
Anthropometrists, field staff, and laboratory staff were blinded to BMI at enrollment (9), maternal height and asset score (21). The binary
the group assignments. The data analyst remained blinded until all outcomes were analyzed using logistic regression first by running the
decisions regarding outliers had been made. unadjusted model, and then the adjusted model using the prespecified
covariates. For all analyses, only covariates significantly associated with
Sample size and data analysis. the outcome at 10% level of significance in a bivariate analysis were
included in the final adjusted analysis, except for child age which was
We hypothesized that (a) children in the LNS group would differ
included in all models. We also tested for interaction with a set of
in height and HAZ compared to children in the combined IFA and
prespecified covariates (maternal height, estimated prepregnancy BMI,
MMN tablet group at 4–6 y, and (b) children in the LNS group
nulliparity at enrollment, total years of schooling, and child sex). The
would not have significantly higher % body fat compared to the
test for interaction was considered significant at P-interaction <0.10.
children in the combined IFA and MMN tablet (non-LNS) group at
We conducted further analysis by stratifying the intervention groups
4–6 y (noninferiority analysis). For the second hypothesis, we chose a
(non-LNS compared with LNS) by categories of the significant effect
noninferiority approach because our objective was to rule out potential
modifier.
adverse effects (i.e., greater body fatness in the LNS group), given
For the noninferiority hypothesis regarding % body fat, ANCOVA
concerns that a high-fat product might lead to greater body fatness,
was used to analyze the data. The difference in the means of the 2 groups
and we prespecified a noninferiority margin of 1% in body fat, based
and associated 95% CI was compared with the predetermined margin
on a SD of 2.6% in fat mass (16) and the aim of ruling out any effect
of inferiority (1% in body fat).
greater than a moderate effect size of 0.4 (noninferiority margin divided
by SD). Noninferiority was established if the difference in the means
and the CI around it were below the predetermined margin of inferiority
(where lower is better). Both hypotheses were based on combining the 2 Results
control groups (IFA andMMN), as there was no evidence of differences
in growth or body composition in other studies comparing MMN Of the 1222 children who were eligible to participate in the
to IFA (17). Nevertheless, a sensitivity analysis was also performed follow-up,we enrolled 662 in the non-LNS group and 352 in the
comparing the 3 groups to check for any differences between the IFA LNS group (Figure 1). By the end of the data collection period
andMMN groups. Primary outcomes were height (cm), height-for-age z (January to December 2016), we were able to measure 79%
score (HAZ) and% FM at 4–6 y. Secondary outcomes included stunting of the eligible children, 623 in the non-LNS group and 338 in
(HAZ < -2), weight (kg), weight-for-age z score (WAZ), underweight
(WAZ -2), overweight (BMIZ +1), arm fat area (AFA), arm muscle the LNS group. The reasons for nonparticipation in the follow-< >
area (AMA), and high body fat. AFA and AMA were calculated using up were inability to trace the mother (n = 96), consent refusal
the following equations: (n= 43), child death after the main trial ended (n= 5), no longer
residing in the study area (n = 64), and loss to follow-up after
AFA(cm2) = (TSF X MUAC 2)− ( X(TSF)2 4) enrollment in the follow-up (n = 53). The proportion lost to/ π /
AMA(cm2) = [(MUAC− (1)(π X TSF))]2 4 follow-up was 23.7% in the non-LNS group and 16.5% in the/ π
LNS group (P = 0.005).
Table 1 shows the background characteristics at enrollment
where TSF and MUAC are in cm (18). into the main trial of women whose children were enrolled into
High body fat was defined as % body fat ≥20. About half the
children in our sample were <5 y of age, and we could not find the follow-up. More than 90% of the women were married
a reference for high body fat for children aged <5 y. There are or cohabiting, a third were nulliparous at baseline, and 10.9%
2
documented sex differences in %FM after 5 y of age (19), but it were underweight (BMI < 18.5 kg/m ). Of the eligible children
is not clear whether a sex difference is generally found in children from the main trial, maternal baseline characteristics for those
aged <5 y, or the magnitude of any such difference. Because of this included in this follow-up (n = 961) were not significantly
850 Kumordzie et al.
TABLE 1 Background characteristics at enrollment into the compared with children in the non-LNS group: the upper end
main trial of women whose children were eligible for follow-up, of the CI for the difference in means between the 2 groups
and children in the International Lipid-Based Nutrient (+0.81% body fat, Table 2) did not exceed our predetermined
Supplements (iLiNS)-DYAD Ghana trial at follow-up1 inferiority margin of 1%.
Non-LNS The binary anthropometric outcomes are presented in
Characteristic (n= 623) LNS (n= 338) Table 3. Overall prevalence of stunting (6.4%) was low, few
children were overweight (2.9%), and 16.1% had high body
Maternal characteristics fat (≥20%). There were no significant differences between
Age, y 26.8± 0.2 26.8± 0.3 intervention groups in any of the binary outcomes in either
Gestational age at enrollment, wk 16.1± 3.2 16.1± 3.3 the 2-group (Table 3) or 3-group (Supplemental Table 3)
Formal education, y 7.6± 3.4 7.6± 3.8 comparisons.
Married or cohabiting, % 93.7 92.6 Adjustment for prespecified covariates did not alter any of
Asset score2 0.07± 0.95 −0.09± 0.983 the above findings (data not shown).
Primiparous women, % 31.9 32.5 We examined potential effect modification by maternal
Weight, kg 61.7± 0.5 62.9± 0.7 height, BMI, primiparity, total years of schooling, and child sex.
Height, cm 158.9± 5.8 159.2± 5.4 We found interactions (P-interaction <0.10, before correction
Prepregnancy BMI4,5, kg/m2 24.5± 4.7 24.9± 4.5 using the Benjamini-Hochberg procedure) of intervention group
Overweight (BMI≥ 25), % 30.4 34.1 with estimated maternal prepregnancy BMI for child height
Underweight (BMI< 18.5), % 13.1 6.9 (Figure 2), HAZ, and % fat mass, and with maternal height
MUAC4, cm 27.8± 4.2 28.3± 4.3 for child % fat mass. For the interactions with maternal
TSF4, mm 18.7± 7.4 19.4± 7.5 prepregnancy BMI, height was greater in the LNS group
Child characteristics compared with the non-LNS group (+1.1 cm; 95% CI: 0.2, 2.1
Age of child at follow-up, y 4.9± 0.6 5.0± 0.6 cm, P= 0.017) among children whose mothers had a BMI< 25
Sex of child, % boys 47.7 48.6 kg/m2 (n = 565) at baseline, but there was no difference (+0.1
1Values are mean ± SD unless otherwise stated. LNS, lipid-based nutrient cm; 95% CI: −1.1, 1.3 cm, P = 0.874) among children of
supplements group; MUAC, midupper arm circumference; non-LNS, iron folic acid overweight mothers (BMI ≥ 25 kg/m2; n = 380) (Figure 2).
group + multiple micronutrients group; TSF, triceps skinfold.
2 A similar trend was observed for child HAZ: +0.19 (95% CI:Household asset score was constructed based on ownership of a set of assets (radio,
television, refrigerator, and stove), lighting source, drinking water supply, sanitation 0.03, 0.35, P = 0.018) in the LNS group among children of
2
facilities, and flooring materials, developed into an index (with a mean of 0 and SD of mothers with BMI < 25 kg/m , and no effect of LNS (−0.10;
1) using principal components analysis. 95% CI −0.03, 0.11, P = 0.349) among children of overweight
3Different from control, P < 0.05. mothers. We did not observe a difference in % fat mass
4Data available for less than the full sample. Sample size for these analyses of 611 between the non-LNS and LNS groups when we stratified by
compared with 334 (non-LNS compared with LNS) for weight, BMI, MUAC, and TSF
because some of the women did not have baseline anthropometric data. maternal BMI (Figure 3), although P-interaction was 0.037. The
5Estimated prepregnancy BMI was calculated from estimated prepregnancy weight interaction of intervention group with maternal height, with
(based on polynomial regression with gestational age, gestational age squared, and regard to child % fat mass (Supplemental Figure 1), was similar
gestational age cubed as predictors) and height at enrollment. to the pattern shown with maternal BMI. However, none of
the P-for-interaction values was significant after correction for
multiple comparisons using the Benjamini/Hochberg procedure.
different from those of children who were not in the follow-
up, except for nulliparity (32.2% among those included in
the follow-up sample compared with 38.3% among those not Discussion
included, P = 0.041, Supplemental Table 1). Among those
included in the follow-up, there were no significant differences In this follow-up of the iLiNS-DYAD Ghana cohort, we did not
in background characteristics between the non-LNS and LNS observe any overall effect of LNS provided to mothers during
groups (Table 1), except in household asset score, which was pregnancy and the first 6 mo postpartum and to their children
higher in the non-LNS group (P = 0.017). At follow-up, the from 6 to 18 mo of age on child height, HAZ, or any of the
mean age (mean ± SD) of the children was 4.9 ± 0.6 y and the other anthropometric measures at 4–6 y of age. Although there
percentage of boys in the sample was 48%, and these did not were significant differences in mean length (0.61 cm; P= 0.001)
differ between groups. and weight (210 g; P = 0.010) between the LNS and non-
The means ± SD for height, HAZ, weight, WAZ, BMIZ, LNS groups at 18 mo (9), these differences were not sustained
and %FM in this cohort were 106.5 ± 5.5 cm, −0.54 ± 0.95, several years later. However, in prespecified subgroup analyses,
16.6 ± 2.2 kg,−0.71 ± 0.86,−0.56 ± 0.83, and 15.4 ± 4.9%, the results suggested that maternal BMI may have modified the
respectively. Table 2 shows the continuous unadjusted anthro- effect of the intervention: among children of women who were
pometric measures and body composition results.At 4–6 y, there not overweight at baseline, child height was +1.1 (95% CI:
were no significant differences between the children in the non- 0.2, 2.1) cm greater in the LNS group compared with the non-
LNS compared with LNS groups for any of these outcomes. LNS group at 4–6 y, whereas there was no difference between
In the 3-group analysis, we also did not find any significant intervention groups among children of overweight women.
overall group differences or pairwise differences between the The noninferiority results support our hypothesis that
IFA and MMN groups (Supplemental Table 2); for the primary provision of LNS in early life did not increase % body fat at 4–
outcome, height, the mean ± SE values were 105.9 ± 0.3 cm, 6 y of age in the overall cohort. There was evidence suggesting
106.6 ± 0.3 cm, 106.7 ± 0.2 cm for the IFA, MMN, and LNS effect modification bymaternal prepregnancy BMI,with slightly
groups, respectively (P = 0.075). greater % body fat in the LNS (compared with non-LNS)
The results for the noninferiority analysis indicate that group among children of women who were not overweight at
children in the LNS group did not have higher % body fat baseline, but slightly lower % body fat in the LNS (compared
Early nutrition supplementation and later growth 851
TABLE 2 Anthropometric and body composition measurements of children in the International Lipid-Based Nutrient Supplements
(iLiNS)-DYAD Ghana trial at 4–6 y1
Ratio of the geometric
Variable Non-LNS (n= 623) LNS (n= 338) P-value Difference in mean (95% CI) mean (95% CI)
Height, cm 106.3± 0.2 106.7± 0.2 0.226 0.36 (−0.23, 0.95) —
Height-for-age z score −0.57± 0.04 −0.49± 0.05 0.226 0.08 (−0.05, 0.21) —
Weight, kg 16.5± 0.1 16.7± 0.1 0.214 0.17 (−0.10, 0.43) —
Weight-for-age z score −0.74± 0.03 −0.67± 0.05 0.252 0.07 (−0.05, 0.18) —
BMI-for-age z score −0.58± 0.03 −0.55± 0.05 0.606 0.03 (−0.08, 0.14) —
MUAC, cm 15.4± 0.1 15.5± 0.1 0.382 0.07 (−0.09, 0.23) —
TSF2, mm 7.0 (6.9, 7.1) 7.1 (6.9, 7.2) 0.392 — 1.01 (0.98, 1.04)
AFA 2, cm2 5.4 (5.3, 5.5) 5.5 (5.3, 5.6) 0.399 — 1.01 (0.98, 1.05)
AMA 2, cm2 13.7 (13.6, 13.9) 13.7 (13.6, 14.0) 0.891 — 1.00 (0.98, 1.02)
FM 3, kg 2.6± 0.0 2.6± 0.1 0.491 0.05 (−0.08, 0.17) —
% FM 3,4 15.3 (15.0, 15.7) 15.5 (14.9, 16.0) 0.630 0.16 (−0.49, 0.81) —
FFM 3, kg 14.0± 0.1 14.1± 0.1 0.389 0.10 (−0.13, 0.34) —
% FFM 3 84.9 (84.4, 85.3) 84.6 (84.0, 85.2) 0.429 −0.29 (−1.02, 0.43) —
1Values represent mean ± SE and the difference in mean (95% CI) unless otherwise stated. Results are based on ANCOVA (SAS PROC GLIMMIX). AFA, arm fat area; AMA, arm
muscle area; FFM, fat-free mass; FM, fat mass; LNS, lipid-based nutrient supplements group; MUAC, midupper arm circumference; non-LNS, iron folic acid group + multiple
micronutrients group; TSF, triceps skinfold.
2Values are geometric means (95% CI) and ratio of the geometric mean (95% CI). n = 622 compared with 337 (non-LNS compared with LNS).
3Values are based on less than the full sample for FM and %FM (599 compared with 326, non-LNS compared with LNS) and for FFM and %FFM (604 compared with 327,
non-LNS compared with LNS) because some participants did not take part in the body composition procedure, there was insufficient sample for analysis, or there were data
collection or lab issues.
4An outlier in the %FM variable was truncated using the value representing the 95th percentile of the distribution to make the distribution more normal.
with non-LNS) group among children of overweight women. who were blinded to group assignment and all measurements
Because the difference between intervention groups was not were done at a central site. Overall, we were able to measure
significant in either of these subgroups, we cannot conclude that 79% of the children born during the main trial who had
there were any positive or negative effects, although Figure 3 not died, with most of the loss to follow-up attributable to
suggests that LNS eliminated the differential between children relocation of participant households. A potential limitation
of nonoverweight and overweight mothers: % body fat in the is the observed differential loss to follow-up between the
LNS group was similar across both subgroups, whereas in the groups, with a higher loss in the non-LNS group. Despite
non-LNS group % body fat was 1.5% higher in the subgroup this differential loss, the groups did not differ in maternal
of children born to overweight mothers. The overall percentage background characteristics (except in asset score which was
of children with high body fat (≥20%) was 16.1%,which is not lower in the LNS group), and adjustment of the statistical
excessive given that the cutoff was based on the 85th percentile models for background characteristics did not change any of our
of the reference population (19). conclusions. In addition, the sample of children included in the
A strength of our study is the use of the deuterium dilution follow-up did not differ in maternal background characteristics
method to determine body composition, although 2 limitations compared to the full sample of women enrolled in the main
of our procedures should be noted: we did not have an trial (except for primiparity), indicating that the results could
objective measure of health before the deuterium dilution be generalizable to the study population.
sample collection, which is desirable because health status To our knowledge, only 2 previous studies included a follow-
influences body water, and the children were not required to fast up assessment of growth of children exposed to a nutritional
because this is difficult for children at this age. Another strength supplement during both prenatal and postnatal life, both of
is that we used trained and standardized anthropometrists which were conducted in the 1970s: the Institute of Nutrition
TABLE 3 Prevalence of underweight, stunting, overweight, and % body fat ≥20% among children in the International Lipid-Based
Nutrient Supplements (iLiNS)-DYAD Ghana trial at 4–6 y1
Non-LNS (n= 622) LNS (n= 338) OR (95% CI) P-value
Stunting (HAZ< -2 SD) — — —
Prevalence, % 6.3 6.5 1.00 (0.58, 1.73) 0.993
Underweight (WAZ< -2SD) — — —
Prevalence, % 5.6 6.2 1.10 (0.63, 1.93) 0.730
Overweight (BMIZ> 1 SD) — — —
Prevalence, % 2.9 3.0 1.10 (0.47, 2.54) 0.827
Body fat≥20% — — —
Prevalence 2, % 14.3 17.4 1.26 (0.87, 1.18) 0.218
1Values are the percentage of participants whose response was “yes” for the outcome in question and OR (95% CI) obtained by comparing the groups. Reference = non-
LNS group for all outcomes. Results are based on logistic regression (SAS PROC GLIMMIX). BMIZ, BMI-for-age z score; HAZ, height-for-age z score; LNS, lipid-based nutrient
supplements group; non-LNS, iron folic acid group + multiple micronutrients group; WAZ, weight-for-age z score.
2Values are based on less than the full sample for overweight (601 compared with 328, non-LNS compared with LNS) because some participants did not take part in the body
composition procedure, there was insufficient sample for analysis, or there were data collection or lab issues.
852 Kumordzie et al.
intervention ended. In the INCAP trial, mean height in the
Atole group during adolescence was 1.2 and 2.1 cm greater for
males and females, respectively, compared to the Fresco group
(the difference was significant only in females). In the earlier
trials, the supplementation period extended from pregnancy
until at least 3 y of age, whereas supplementation in our trial
ended at 18 mo. The types of supplement provided also differed
substantially among trials. We provided small-quantity LNS,
which provided a similar amount of energy per serving as
Atole but less protein. The INCAP trial provided a low-energy
fortified supplement to mothers and children in the control
group, whereas we provided IFA or MMN to mothers only.
The Bogotá trial provided 623 kcal and 30 g protein to each
family member and 856 kcal and 38 g protein to the mother
daily from week 26 of pregnancy until 36 mo postpartum; the
FIGURE 2 Child height at 4–6 y by intervention group (LNS control group received no supplement. The supplements in the
compared with non-LNS), stratified by maternal BMI at enrollment Guatemala and Bogotá trials provided more milk than the LNS:
into the International Lipid-Based Nutrient Supplements (iLiNS)-DYAD at least 20 g of dry milk per serving of Atole in Guatemala
Ghana trial. Values represent mean (95% CI) from an ANCOVA (22) and 60 g/d of dry skimmed milk for each household
model (SAS PROC GLIMMIX). P-interaction between estimated
maternal prepregnancy BMI as a continuous variable and intervention member in Bogotá (8), compared to 4.8 g/d of dry milk
group = 0.046, before Benjamini/Hochberg correction, adjusting for from LNS (11). Milk and milk products have been associated
maternal height, years of education, nulliparity, asset score, and child with increased linear growth, although the exact mechanism
sex. LNS, lipid-based nutrient supplements group; non-LNS, iron-folic is not known (23). Another difference between these earlier
acid group + multiple micronutrients group. trials and the iLiNS-DYAD Ghana trial is that malnutrition
was greater in Guatemala and Colombia at that time, when
these lower income countries had a high disease burden, poor
of Central America and Panama (INCAP) longitudinal study in access to health care, and poor sanitation, whereas our study
Guatemala (7) and the Bogotá study of Malnutrition, Diarrheal population in Ghana had good access to health care and potable
Disease and Child Development in Colombia (8). In both water, most of the population had access to improved toilet
studies, children in the intervention group were taller and facilities, and the prevalence of overweight among the mothers
weighed more than those in the control group at the end of at enrollment was high (38.5%). This may explain in part why
the supplementation period, as was the case in our study as we saw long-term effects on height only in the group of children
well. At follow-up, both of the early trials reported sustained born to women who were not overweight. Among children of
effects. Although we did not find a significant overall difference overweight women, there is likely to be less constraint on the
between the non-LNS and LNS groups at 4–6 y, there was a growth of the child because of maternal nutrition (especially in
nonsignificant difference of 0.8 cm in height between the IFA utero), and thus little potential to benefit from LNS, whereas
and LNS groups, similar to the ∼ 0.6 cm difference between children of nonoverweight mothers may have more potential
these 2 groups at 18 mo (9), and a significant+1.1 cm difference to benefit from nutritional supplementation as evidenced in a
in height between the non-LNS and LNS groups among children vulnerable Bangladeshi population (24).
of women who were not overweight. The Bogotá trial showed Apart from the 2 trials described above, there have been
a difference in height of 2.3 cm at 6 y of age, 3 y after the several follow-up studies of food or nutrient supplementation
during pregnancy only. A 2015 review and meta-analysis of
energy-protein supplementation trials conducted between 1973
and 2014 (25) found no sustained effects of prenatal energy-
protein supplementation alone on child growth status at 12 mo
of age (26–28) or on height, weight, and body fat at 11–17 y
(29).
In another systematic review and meta-analysis of follow-
up studies of prenatal micronutrient supplementation trials, the
meta-analysis showed no difference between groups receiving
prenatal MMNs compared with iron (60 mg) and folic acid
in child HAZ or WAZ (17). Two of the trials mentioned, 1 in
Burkina Faso (30) and the other in Nepal (31, 32) both reported
greater WAZ in the MMN groups at younger ages (1 and 2.5 y),
but these differences were not sustained at older ages (2.5 and
8.5 y, respectively). In a separate trial in Nepal (33), there was an
FIGURE 3 Child % fat mass at 4–6 y by intervention group (LNS increase in mean height in 1 of the intervention groups (vitamin
compared with non-LNS), stratified by maternal BMI at enrollment
into the International Lipid-Based Nutrient Supplements (iLiNS)-DYAD A + iron + folic acid + zinc) compared to the control group
Ghana trial. Values represent mean (95% CI) from an ANCOVA (vitamin A alone), but this difference was not evident in the
model (SAS PROC GLIMMIX). P-interaction between estimated group whose mothers received multiple (>4) micronutrients.
maternal prepregnancy BMI as a continuous variable and intervention The 2 Nepal trials – Janakpur (31, 32) and Sarlahi (33) – and 1
group = 0.037, before Benjamini/Hochberg correction, adjusting for trial in Bangladesh, MINIMat (34), included measurement of
nulliparity and child sex. LNS, lipid-based nutrient supplements group; body composition. The MINIMat trial showed no difference
non-LNS, iron-folic acid group + multiple micronutrients group. in skinfold thicknesses, whereas the Sarlahi Nepal trial showed
Early nutrition supplementation and later growth 853
lower skinfold thickness in 1 intervention group (iron + folic 8. Super CM, Herrera MG, Mora JO. Long-term effects of food
acid + zinc) compared to controls at 7.5 y, and the Janakpur supplementation and psychosocial intervention on the physical growth
Nepal trial showed higher TSF in the MMN group at 2.5 y but of Colombian infants at risk of malnutrition. Child Dev 1990;61(1):
29–49.
not later. There was no reported difference in lean mass or fat
9. Adu-Afarwuah S, Lartey A, Okronipa H, Ashorn P, Peerson JM,
mass in either the MINIMat or Janakpur Nepal studies. These Arimond M, Ashorn U, Zeilani M, Vosti S, Dewey KG. Small-
findings are consistent with our findings from the follow-up of quantity, lipid-based nutrient supplements provided to women during
the iLiNS-DYAD Ghana trial of no differences between the IFA pregnancy and 6 mo postpartum and to their infants from 6 mo of age
and MMN groups. increase the mean attained length of 18-mo-old children in semi-urban
We conclude that there was no overall effect of LNS on Ghana: a randomized controlled trial. Am J Clin Nutr 2016;104(3):797–808.
height or HAZ at 4–6 y of age in this cohort, which had 10. Adu-Afarwuah S, Lartey A, Okronipa H, Ashorn P, Zeilani M, Peerson
a low stunting rate, but the results suggested an increase JM, Arimond M, Vosti S, Dewey KG. Lipid-based nutrient supplement
in height and HAZ in the LNS group among children of increases the birth size of infants of primiparous women in Ghana. Am
nonoverweight women. Although this finding needs to be J Clin Nutr 2015;101:834–46.
confirmed in similar populations, it has potential implications 11. Arimond M, Zeilani M, Jungjohann S, Brown KH, Ashorn P,
for targeting interventions such as LNS. There was no adverse Allen LH, Dewey KG. Considerations in developing lipid-based
nutrient supplements for prevention of undernutrition: experience from
impact of LNS on child body fatness, suggesting that a small the International Lipid-Based Nutrient Supplements (iLiNS) Project.
daily quantity of a high-fat fortified supplement in early life does Matern Child Nutr 2015;11(Suppl 4):31–61.
not contribute to child overweight. 12. AGENCY IAE. Introduction to Body Composition Assessment Using
the Deuterium Dilution Technique with Analysis of Saliva Samples by
Acknowledgments Fourier Transform Infrared Spectrometry. Vienna: International Atomic
We would like to thank Sophia Amenya, Florence Dokyi, Energy Agency; 2011.
Frederick Yeboah, and Elizabeth Amue for their help with 13. Brunette W, Sundt M, Dell N, Chaudhri R, Breit N, Borriello G. Open
the deuterium dilution saliva collection and anthropometry, data kit 2.0: expanding and refining information services for developing
regions. Proceedings of the 14th Workshop on Mobile Computing
Solace M. Tamakloe for coordinating the field staff, and Systems and Applications; Jekyll Island, Georgia. 2444790: ACM;
Edward Christian Brown, Yaa Pokuaa Akomea, and Kennedy 2013. p. 1–6.
Amewosina of the Nutrition Research Center, GAEC for their 14. de Onis M, Onyango AW, Van den Broeck J, Chumlea WC, Martorell
help with the laboratory analysis. We would also like to R.Measurement and standardization protocols for anthropometry used
acknowledge Dr.Anna Lartey who supervised the iLiNS-DYAD- in the construction of a new international growth reference. Food Nutr
Ghana trial. Bull 2004;25(1 Suppl):S27–36.
The authors’ responsibilities were as follows—SMK, SA- 15. WHO Multicentre Growth Reference Study Group. WHO ChildGrowth Standards based on length/height,weight and age.Acta Paediatr
A, MA, RRY, MEO, HO, ELP, BMO, and KGD: designed the Suppl 2006;450:76–85.
study; SMK, SA-A, and MEO: conducted the research; TA and 16. Prins M, Hawkesworth S, Wright A, Fulford AJ, Jarjou LM, Prentice
RB: conducted the training on the deuterium dilution method AM, Moore SE. Use of bioelectrical impedance analysis to assess
and supervised the laboratory analysis; RRY: was the study body composition in rural Gambian children. Eur J Clin Nutr
statistician; RRY and SMK: performed the statistical analysis; 2008;62(9):1065–74.
RRY and KGD: advised on the analysis; SMK and KGD: 17. Devakumar D, Fall CH, Sachdev HS, Margetts BM, Osmond C, Wells
JC, Costello A, Osrin D. Maternal antenatal multiple micronutrient
wrote the manuscript; SA-A, MA, RRY, MEO, HO, ELP, and supplementation for long-term health benefits in children: a systematic
BMO: reviewed the draft manuscript; and all authors: read and review and meta-analysis. BMC Med 2016;14:90.
approved the final paper. 18. Gibson RS. Principles of Nutritional Assessment: Oxford University
Press; 2005.
19. McCarthy HD,Cole TJ, Fry T, Jebb SA, Prentice AM.Body fat reference
curves for children. Int J Obes 2006;30(4):598–602.
References 20. Cohen J. Statistical power analysis for the behavioral sciences. 2nd ed.
Hillsdale, NJ, USA: Lawrence Earlbaum Associates, Inc.; 1988.
1. Black RE, Allen LH, Bhutta ZA, Caulfield LE, de Onis M,
Ezzati M, Mathers C, Rivera J. Maternal and child undernutrition: 21. Vyas S, Kumaranayake L. Constructing socio-economic status indices:
global and regional exposures and health consequences. Lancet how to use principal components analysis. Health Policy Plan
2008;371(9608):243–60. 2006;21(6):459–68.
2. Black RE, Victora CG, Walker SP, Bhutta ZA, Christian P, de Onis M, 22. Martorell R, Habicht JP, Rivera JA. History and design of the INCAP
Ezzati M, Grantham-McGregor S, Katz J, Martorell R, et al. Maternal longitudinal study (1969–77) and its follow-up (1988–89). J Nutr
and child undernutrition and overweight in low-income and middle- 1995;125(4 Suppl):1027s–41s.
income countries. Lancet 2013;382(9890):427–51. 23. Yackobovitch-GavanM, PhillipM,Gat-Yablonski G.HowMilk and Its
Proteins Affect Growth, Bone Health, and Weight. Horm Res Paediatr
3. Schwarzenberg SJ, Georgieff MK; COMMITTEE ON NUTRITION. 2017;88(1):63–9.
Advocacy for improving nutrition in the first 1000 days to support
childhood development and adult health. Pediatrics 2018 141(2). pii: 24. Persson LA, Arifeen S, Ekstrom EC, Rasmussen KM, Frongillo
e20173716. doi: 10.1542/peds.2017-3716. EA, Yunus M. Effects of prenatal micronutrient and early food
supplementation on maternal hemoglobin, birth weight, and infant
4. Gluckman PD,HansonMA,Pinal C.The developmental origins of adult mortality among children in Bangladesh: the MINIMat randomized
disease. Matern Child Nutr 2005;1(3):130–41. trial. JAMA 2012;307(19):2050–9.
5. Dewey KG, Begum K. Long-term consequences of stunting in early life. 25. Ota E, Hori H, Mori R, Tobe-Gai R, Farrar D. Antenatal dietary
Matern Child Nutr 2011;7:5–18. education and supplementation to increase energy and protein intake.
6. Corvalan C, Gregory CO, Ramirez-Zea M, Martorell R, Stein AD. Cochrane Database Syst Rev 2015(6):Cd000032.
Size at birth, infant, early and later childhood growth and adult 26. Blackwell RQ, Chow BF, Chinn KS, Blackwell B-N, Hsu S.
body composition: a prospective study in a stunted population. Int J Prospective maternal nutrition study in taiwan-rationale, study
Epidemiol 2007;36(3):550–7. design, feasibility, and preliminary findings. Nutr Rep Int 1973;7(5):
7. Rivera JA, Martorell R, Ruel MT, Habicht JP, Haas JD. Nutritional 517–32.
supplementation during the preschool years influences body size 27. Huybregts L, Roberfroid D, Lanou H, Menten J, Meda N, Van
and composition of Guatemalan adolescents. J Nutr 1995;125(4 Camp J, Kolsteren P. Prenatal food supplementation fortified
Suppl):1068S–77S. with multiple micronutrients increases birth length: a randomized
854 Kumordzie et al.
controlled trial in rural Burkina Faso. Am J Clin Nutr 2009;90(6): mid-childhood in Nepal: follow-up of a double-blind randomised
1593–600. controlled trial. Lancet Glob Health 2014;2(11):e654–63.
28. Rush D, Stein Z, Susser M. A randomized controlled trial of 32. Vaidya A, Saville N, Shrestha BP, Costello AM, Manandhar DS, Osrin
prenatal nutritional supplementation in New York City. Pediatrics D. Effects of antenatal multiple micronutrient supplementation on
1980;65(4):683–97. children’s weight and size at 2 years of age in Nepal: follow-up of
29. Hawkesworth S, Prentice AM, Fulford AJ, Moore SE. Dietary a double-blind randomised controlled trial. Lancet 2008;371(9611):
supplementation of rural Gambian women during pregnancy does not 492–9.
affect body composition in offspring at 11–17 years of age. J Nutr 33. Stewart CP, Christian P, LeClerq SC,West KP, Jr., Khatry SK. Antenatal
2008;138(12):2468–73. supplementation with folic acid + iron + zinc improves linear growth
30. Roberfroid D, Huybregts L, Lanou H, Ouedraogo L, Henry MC,Meda and reduces peripheral adiposity in school-age children in rural Nepal.
N, Kolsteren P. Impact of prenatal multiple micronutrients on survival Am J Clin Nutr 2009;90(1):132–40.
and growth during infancy: a randomized controlled trial. Am J Clin 34. Khan AI, Kabir I, Hawkesworth S, Ekstrom EC, Arifeen S, Frongillo
Nutr 2012;95(4):916–24. EA, Persson LA. Early invitation to food and/or multiple micronutrient
31. Devakumar D, Chaube SS,Wells JC, Saville NM, Ayres JG,Manandhar supplementation in pregnancy does not affect body composition in
DS, Costello A, Osrin D. Effect of antenatal multiple micronutrient offspring at 54 months: follow-up of the MINIMat randomised trial,
supplementation on anthropometry and blood pressure in Bangladesh. Matern Child Nutr 2015;11(3):385–97.
Early nutrition supplementation and later growth 855