Small-quantity lipid-based nutrient supplements for prevention of child malnutrition and promotion of healthy development: Overview of individual participant data meta-analysis and programmatic implications

Small-quantity lipid-based nutrient supplements (SQ-LNS) were designed to provide multiple micronutrients within a food base that also provides energy, protein and essential fatty acids, and were targeted towards the prevention of malnutrition in low- and middle-income countries. Previous meta-analyses have demonstrated beneficial effects of SQ-LNS on child growth, anemia and mortality. To further examine efficacy and effectiveness of SQ-LNS, and explore study-level and individual-level modifiers of the effects, we conducted an individual participant data (IPD) meta-analysis of 14 randomized controlled trials of SQ-LNS provided to infants and young children 6 to 24 mo of age (n > 37,000). We examined growth, development, anemia and micronutrient status outcomes. Children who received SQ-LNS had a 12-14% lower prevalence of stunting, wasting and underweight, were 16-19% less likely to score in the lowest decile for language, social-emotional, and motor development, and had a 16% lower prevalence of anemia and 64% lower prevalence of iron-deficiency anemia, compared to control group children. For most outcomes, beneficial effects of SQ-LNS were evident regardless of study-level characteristics including region, stunting burden, malaria prevalence, sanitation, water quality, duration of supplementation, frequency of contact or average reported compliance with SQ-LNS. For certain outcomes, targeting based on population-level socioeconomic status or undernutrition may be worthwhile, as the benefits of SQ-LNS for iron status, anemia and child development were larger in sub-groups with a greater potential to benefit. A greater impact of SQ-LNS might be possible by co-packaging it with interventions that reduce constraints on response, such as prevention and control of pre-and postnatal infections, improving maternal nutrition, improving health care access, and promotion of early child development. Policy-makers and program planners should consider including SQ-LNS in strategies to reduce child mortality, stunting, wasting, anemia, iron deficiency and developmental impairments. This study was registered at www.crd.york.ac.uk/PROSPERO as CRD42019146592, CRD42020159971 and CRD42020156663.


Introduction
The first trials evaluating SQ-LNS were published in 2007-2008, and results were promising 110 with regard to prevention of linear growth faltering in Ghana (29) and of severe stunting in 111 Malawi (30). Subsequently, the International Lipid-based Nutrient Supplements (iLiNS) Project  137 Overview of methods 138 The IPD meta-analysis presented in the accompanying articles is based on pooled data from 14 139 randomized controlled trials of SQ-LNS. This work followed best practices for pre-registration, 140 transparency, and reproducibility, with protocols and statistical analysis plans posted online 141 (osf.io/ymsfu and at PROSPERO CRD42019146592, CRD42020159971 and CRD42020156663) 142 (31)(32)(33)(34). The trials were identified beginning with those published in a recent Cochrane Review 143 (21), and supplemented with additional studies identified through a systematic review of studies 144 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) shown in Box 1 were invited to contribute data and participate in the analysis. with other non-nutritional interventions (i.e., water, sanitation and hygiene (WASH)), all arms 152 that provided SQ-LNS were combined into one group. All non-LNS arms were combined into a 153 single comparator ("control") group for each trial [excluding intervention arms that received 154 non-LNS child supplementation, e.g., MNP, fortified-blended food]. For trials that provided both 155 maternal and child LNS, we conducted analyses both with ("all-trials analysis") and without 156 ("child-LNS-only") the maternal LNS arms. The main effects did not differ between these two 157 analyses by more than 20% for continuous outcomes or by 0.05 for prevalence ratios, so the 158 results of the all-trials analyses are presented as the principal findings.

160
We conducted several pre-specified sensitivity analyses: 1) separate comparisons within multi- 167 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) Three types of statistical analyses were conducted to separately investigate 1) full sample 168 treatment effects, 2) effect modification by study-level characteristics, and 3) effect modification 169 by individual-level characteristics. For all 3 sets of analyses, we used a two-stage approach 170 following a complete-case intention-to-treat framework. We used bivariate meta-regression 171 modeling to examine effect modification by study-level characteristics. We modeled effect 172 modification by individual-level characteristics within each study, and then pooled the estimates.

173
Potential effect modifiers examined are shown in Table 1.

192
Six trials were conducted within existing community-based or clinic-based programs (36,39,41,193 43, [46][47][48]; in the other trials, all activities were conducted by research teams. Seven trials 194 provided minimal messaging on IYCF other than reinforcing the normal IYCF messages already 195 promoted in that setting (29,36,38,40,41,44,45), and 7 trials provided expanded SBCC on 196 IYCF that went beyond the usual messaging, either in just the SQ-LNS intervention arms (37, 197 39, 42, 47, 48) or in all arms including the non-SQ-LNS control arm (35,43,46). Three trials 198 included arms with WASH interventions (37,42,47,48). Most trials included an active control 199 arm (i.e., similar contact frequency as for intervention arms) but 3 included only a passive 200 control arm (29,37,38).  Overall, when combining data from all of the trials, we found significant effects of SQ-LNS 211 across all three outcome domains ( Table 3). Children who received SQ-LNS had a 12-14% 212 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted February 18, 2021. ; https://doi.org/10.1101/2021.02.15.21251449 doi: medRxiv preprint lower prevalence of stunting, wasting and underweight (23), were 16-19% less likely to score in 213 the lowest decile for language, social-emotional, and motor development (24), and had a 16% 214 lower prevalence of anemia and 64% lower prevalence of iron-deficiency anemia (25), compared 215 to control children who did not receive SQ-LNS. These findings add to those of a recently 216 published meta-analysis of many of these same trials reporting a 27% lower risk of mortality 217 between 6 and 24 mo of age (22).  The overall effects of SQ-LNS on the growth and development outcomes were modest.

233
However, the effects were generally more consistent, and for some outcomes more substantial, 234 compared to other nutrition interventions for children under 2 y of age, such as nutrition 235 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 18, 2021. ; https://doi.org/10.1101/2021.02.15.21251449 doi: medRxiv preprint education, micronutrient supplementation or fortification, and fortified blended foods (50)(51)(52)(53)(54). 236 Many studies have examined the impacts of behavior change interventions focused on improving 237 complementary feeding practices. A 2018 Cochrane review concluded that such interventions are 238 effective at improving reported feeding practices, but there was insufficient evidence to draw 239 conclusions with respect to effects on growth, development, anemia, or micronutrient status (51). Effects of SQ-LNS on anemia and iron deficiency were similar to those reported in a recent 247 meta-analysis of MNP (18% reduction in anemia and 53% reduction in iron deficiency; (55)).

248
However, that review demonstrated no effects of MNP on child growth and there was 249 insufficient evidence to evaluate effects on child development or mortality. the prevalence ratio between the 2 groups for a binary outcome (e.g., the relative risk of 268 stunting), and as the difference in absolute prevalence between the 2 groups for a binary outcome 269 (e.g., the percentage point difference in prevalence of stunting). We considered all 3 types of 270 outcomes together and in reference to 3 different theoretical scenarios: potential to benefit, 271 potential to respond, and cutoff effects.

272
Potential to benefit: One possibility is that certain subgroups of children may be more 273 likely to benefit from the SQ-LNS intervention, perhaps due to greater nutritional deficits at 274 baseline (14). This is illustrated in Figure 1A. The children with greater nutritional or 275 developmental deficits in subgroup a-1 are shown here as being slightly smaller and thinner 276 compared to the children in subgroup a-2. When SQ-LNS is provided, the children in subgroup 277 a-1 may benefit more. Our interpretation in these cases is that there were greater benefits to those 278 who needed it most.  individual-level characteristics, but the characteristics that appeared to be important varied 305 considerably depending on the type of outcome. In the paragraphs that follow, we discuss effect 306 modifiers that were common across multiple outcome domains or that might have programmatic 307 implications, using the concepts of potential to benefit and potential to respond described above 308 to interpret the findings.  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 18, 2021. ; https://doi.org/10.1101/2021.02.15.21251449 doi: medRxiv preprint social-emotional, and motor development scores in low-SES households were +0.06 to +0.12 328 SD, compared to little to no effects among children in higher SES households. Similarly, in sites 329 with a high stunting burden, language, social-emotional, and motor development scores were 330 +0.08 to +0.13 SD higher in the SQ-LNS vs. control groups; in contrast, there were no effects on 331 these outcomes in settings with a low stunting burden. While these differences were apparent for 332 the development outcomes, SES and stunting burden generally did not significantly modify the 333 effects of SQ-LNS on growth, anemia, or iron status outcomes.  Maternal education and depressive symptoms also appeared to modify the effect of SQ-LNS on 366 growth and development, but in seemingly disparate ways. Among children whose mothers had 367 higher education or less depressive symptoms, there were greater effects of SQ-LNS on growth.

368
This may reflect a greater potential to respond to the intervention. Such mothers may have had 369 greater autonomy and agency, and therefore may have been better able to adhere to advice 370 regarding the recommended frequency or dosage of supplementation. By contrast, among 371 children whose mothers had lower education, there were larger beneficial effects of SQ-LNS on 372 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted February 18, 2021 (59), and therefore these children may have lagged behind their peers and 374 had greater room for improvement, with greater potential to benefit from an intervention. longitudinally across a full calendar year. It is also important to note that the data included in the 389 IPD meta-analysis were generally based on cross-sectional outcome assessments, which do not 390 capture the multiple episodes of wasting/acute malnutrition, micronutrient deficiency, or anemia 391 that may occur throughout the study period. In the PROMIS study in Mali, for example, the 392 longitudinal prevalence of acute malnutrition was much higher than the cross-sectional 393 prevalence, and the SQ-LNS intervention reduced the former by 29% but had no significant 394 impact on the latter (46).

395
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  In summary, the evidence suggests that there are important benefits of SQ-LNS for child 405 survival, growth, anemia, iron status, and child development. It must be noted that SQ-LNS is 406 not a stand-alone intervention, as it should always be accompanied by messaging to reinforce 407 IYCF recommendations, including a diverse diet with healthy foods from the key food groups.

408
Nonetheless, it appears to play a protective role when access to nutrient-rich foods is limited for    CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Box 1. Eligibility criteria for the SQ-LNS individual participant data (IPD) analyses
Inclusion criteria • Randomized controlled trial • Conducted in low-or middle-income country • SQ-LNS (< ~125 kcal/d) provided to the intervention group for at least 3 mo between 6 and 24 mo of age • At least one trial group did not receive SQ-LNS or other type of child supplementation • Longitudinal follow-up of each child, or repeated cross-sectional data collection • Individual children eligible for IPD analysis if: o Age at baseline allowed receipt of intervention (supplementation or control group components) for at least 3 mo between 6 and 24 mo of age. o For anemia & micronutrient status outcomes, blood samples were collected during the supplementation period or within 3 mo after the study-defined end of supplementation.
Exclusion criteria • LNS was used for treatment, not prevention, of malnutrition (i.e., only children with severe or moderate malnutrition were eligible for the study) • Conducted in a hospitalized population or among children with a pre-existing disease • SQ-LNS provision was combined with additional supplemental food or nutrients within a single arm (e.g. SQ-LNS + food rations vs. control), and there was no appropriate comparison group that would allow isolation of the SQ-LNS effect (e.g., food rations alone) . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)
B: Subgroup b-1 has a greater potential to respond to a nutritional intervention, due to fewer constraints, as compared to subgroup b-2. SQ-LNS, small-quantity lipid-based nutrient supplements. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)