Our search strategy was developed after first reviewing the papers published in the Lancet 2013 Maternal and Child Nutrition series ^{3, 24} , inclusive of the umbrella review by Bhutta and colleagues ⁷ , for an overview of the literature. Specifically, we hand-searched the bibliography of Bhutta et al. ⁷ for relevant systematic reviews, global health guidelines, and LMIC-based trials. We also performed additional searches in PubMed and the Cochrane Database of Systematic Reviews for more recent trials and other reviews published after 2013. The list of published reviews relevant to this study is provided in Table 1.

For our systematic literature search, we scanned the following databases from inception to August 28, 2019: the Cochrane Central Register of Controlled Trials, Embase, and MEDLINE ( Extended data, Supplementary Tables 1–3) ²⁵ . To increase the sensitivity of our search, we complemented our database searches with relevant trials identified from bibliographies of prior reviews. Table 3 describes the Population, Intervention, Comparator, Outcome, and Study Design (PICOS) criteria used to guide the study selection for our systematic literature review. We included randomized clinical trials on interventions of the following domains: Micronutrient supplements; Food supplements; Deworming for mothers; Maternal and breastfeeding education, and promotion; WASH; and kangaroo care (i.e. skin-to-skin care). The outcomes of interest included change in LAZ, proportions of participants with stunting (defined by LAZ below -2SD), change in length, and change in head circumference. For all intervention domains, except for kangaroo care, we excluded studies that did not report the effects of their respective interventions for at least three months. For kangaroo care, there was no restriction for time of follow-up given short duration nature of this intervention. We excluded non-English language studies.

A team of four reviewers (JJHP, ES, LD, and RM) independently reviewed all abstracts and proceedings identified in the literature searches. The same team independently conducted relevant full-text reviews of relevant papers. If any discrepancies occurred between the studies selected by the same reviewers, a third investigator (KT) provided arbitration.

Using a standardized data sheet in Microsoft Excel, four investigators (JJHP, VJ, NEZ, and HG) independently extracted data for study characteristics, interventions used, patient characteristics at baseline, and outcomes from the final list of selected eligible studies. Any discrepancies observed during data extraction were resolved through discussion between the investigators until consensus was reached.

When sufficient data was available for quantitative assessment, a network meta-analysis or pairwise meta-analysis approach was applied. For all domains of interventions except for kangaroo care, we performed a network meta-analysis for LAZ and stunting. There was a limited number of studies that reported on length and head circumference, so we qualitatively synthesized findings from these trials as an alternative to quantitative analysis. We did not consider kangaroo care as part of the network meta-analysis since these trials involved a shorter intervention duration and follow-up (median follow-up of 2 weeks).

We performed a network meta-analysis within the Bayesian framework in R using the R2WinBUGS v14 package ^{26, 27} . Bayesian models were performed according to the National Institute for Health and Care Excellence (NICE) in their Technical Support Document 2 (TSD2) ²⁸ . Estimates of comparative effectiveness were measured using mean differences in LAZ with the associated 95% credible intervals (95% CrI). In all models, we used an empirically informative heterogeneity prior distribution, as suggested by Rhodes et al. 2016 ²⁹ for LAZ and Turner et al. 2015 ³⁰ for stunting. This was done to stabilize the estimation of heterogeneity in the face of low number of trials per comparison in the network. Our model selection was informed by using the deviance information criterion and the deviance-leverage plots that could help identify outlier(s) in terms of model fit, in accordance with the NICE TSD2 recommendations ²⁸ .

For our primary network meta-analysis, we included both cluster and non-cluster randomized clinical trials (with the unit of randomization set at the individual level). To adjust for clustering effects of the cluster trials, we assumed a conservative intra-cluster correlation coefficient (ICC) of 0.05, and we inflated variances accordingly for continuous outcomes and adjusted the sample sizes and the number of cases for dichotomous outcomes, as recommended by Uhlmann et al. ³¹ We performed a sensitivity analysis by excluding cluster randomized clinical trials in our network meta-analysis. For our pairwise meta-analysis on kangaroo care, we performed a random-effects model using the Metafor R package (in R2WinBUGS v14) ³² . For our network meta-analysis, the estimates of effectiveness were measured using mean differences or relative risk with accompanying 95% credible intervals (CrIs). The estimates of effectiveness were measured using mean differences with accompanying 95% confidence intervals (CIs) for our pairwise meta-analysis on kangaroo care. As no kangaroo care trials involved cluster randomization, our pairwise meta-analysis did not need to adjust for the clustering effect.

Each full text article was evaluated for reporting quality according to the Cochrane Risk of Bias Tool ³³ . The risk of bias assessment within and across studies are provided in the Extended data (Supplementary Table 8) ²⁵ .

The LAZ network ( Extended data, Supplementary Figure 1) ²⁵ included 18 trials consisting of 27,896 mother-infant dyads randomized to 52 intervention arms. The results of our primary analysis on LAZ that included both cluster and non-cluster randomized clinical trials are illustrated in Figure 3. Among micronutrient supplements, multiple micronutrients supplementation (MMN) provided to infants improved LAZ relative to standard-of-care (MMN-C, mean difference: 0.20, 95% CrI: 0.03, 0.35), whereas supplementing breasting mothers with MMN did not improve LAZ (MMN-M, mean difference: -0.02, 95% CrI: -0.18, 0.13). Compared to standard of care, other micronutrient supplements to infants, such as zinc 5 mg (zinc 5 mg C) showed a trend towards improved LAZ, but its CrIs overlapped the null of effect of 0.00 (Mean difference: 0.13; 95% CrI: -0.02, 0.24). Also, other micronutrients to breastfeeding mothers, such as iron and folic acid (IFA-M, mean difference: 0.05, 95% CrI: -0.15, 0.22) and vitamin D (Vit D-M: mean difference: 0.08, 95% CrI: -0.11, 0.26), did not improve LAZ in comparison to standard of care. Similarly, both food supplements and maternal education interventions did not improve LAZ; for instance, in comparison to standard-of-care, combination of IFA and 118 kcal of lipid-based nutrient supplements (IFA+LNS 118 kcal-M) showed a mean difference of 0.08 cm (95% CrI: -0.12, 0.29) for LAZ, where maternal education showed a mean difference of 0.05 cm (95% CrI: -0.12, 0.22 cm). No deworming or WASH interventions showed improvements on LAZ.

The stunting network ( Extended data, Supplementary Figure 3) ²⁵ included 18 trials that consisted of 27,896 mother-infant dyads randomized to 52 intervention arms. The results of our primary analysis that included both cluster and non-cluster randomized clinical trials are illustrated in Extended data, Supplementary Figure 9 ²⁵ . While supplementations of zinc to infants (Zinc 5 mg-C, relative risk [RR]: 0.82, 95% CrI: 0.56, 1.18) and local food supplement within the caloric range of 185–260 kcal fortified with MMN for mothers (Local food 185–260 kcal + MMN M, RR: 0.85, 95% CrI: 0.46, 1.46) showed a trend towards reduced risk of stunting, their CrIs contained the null effect of 1.00. In fact, no interventions demonstrated any improvements towards reducing the risk of stunting.

Our sensitivity analyses were limited to individually (non-cluster) randomized clinical trials only. The network diagrams for LAZ ( Extended data, Supplementary Figure 2) ²⁵ and stunting ( Extended data, Supplementary Figure 4) ²⁵ can be found online along with forest plots ( Extended data, Supplementary Figures 10 and 11) ²⁵ and cross-tables ( Extended data, Supplementary Table 9 and 10) ²⁵ . In our sensitivity analysis on LAZ, no interventions showed improvements for LAZ when compared to standard-of-care, and similarly for stunting, no interventions showed reduced risks for stunting. This is likely due to very few studies being available for sensitivity analyses; only nine trials were available for LAZ and stunting analyses.

Five randomized clinical trials investigating the effects of kangaroo care on linear growth of newborns were included in the pairwise meta-analysis ^{41, 45, 48, 55, 69} . The outcome reporting of these kangaroo care trials was limited to growth velocity of head circumference and length (cm per week). All kangaroo care trials were conducted in Southeastern Asian countries (i.e. India, Malaysia, and Nepal), in hospital settings and involved low birthweight neonates. Kangaroo care consisted of skin-to-skin contact between the mothers’ breasts, where infants in the control group were kept under either a warmer or incubator. The effects of kangaroo care on head circumference and length growth velocities are shown in Extended data, Supplementary Figures 12 and 13 ²⁵ , respectively. All studies except for Acharya 2014 ⁴¹ showed improvements in head circumference (Mean (SD): 31.5 (1.4) 95%CI -0.5, 0.6 ⁴⁵ ; Kangaroo mother care (KMC): 0.75 cm vs conventional method of care (CMC): 0.49 cm p<0.001 ⁵⁵ ;) and length (KMC: 0.99 cm vs CMC 0.70 cm p<0.001 ⁵⁵ ). The pooled estimates of growth velocities for head circumference and length showed improvements for kangaroo care in comparison to the control. Relative to the control, kangaroo care showed an improved mean difference of 0.20 cm/week (95% CI: 0.09, 0.31 cm/week) for head circumference, and for length, a mean difference of 0.23 cm/week (95% CI: 0.10, 0.35 cm/week).

There were twelve trials available for our qualitative summary ^{37, 38, 40, 47, 51, 53, 56, 60, 63– 65, 71} . Of these trials, three were cluster randomized clinical trials that investigated interventions related to maternal education and breastfeeding promotion: Le Roux ⁵⁶ Vazir ⁷¹ , and PROMISE EBF ⁶⁵ did not find differences in their maternal education and breastfeeding promotion interventions. In this three-arm trial conducted in India, mothers in the Complementary Feeding group (n = 202; 20 clusters) received nutrition education messages on breastfeeding and complementary feeding from CHWs, and the mothers in the Complementary Feeding + Play group (n = 195; 20 clusters) received messages on psychosocial stimulation in addition to the same nutritional messages received by the women in the complementary feeding group (the control group received local standard of care; n = 202; 20 clusters). The mothers were approached by the trial investigators during pregnancy, and the interventions began when their child was three months old. By the age of six months, this trial found no differences in terms of length between the three groups (Mean ± SD: Control group: 64.2 ± 2.3; Complementary Feeding group: 64.4 ± 2.5 cm; and Complementary Feeding + Play group: 64.2 ± 2.3).

Additionally, nine trials investigated the effect of nutritional interventions (four trials on micronutrient supplements, three on food supplements, one on both, and one other for deworming) on the incidence of changes in head circumference, or changes in length ^{37, 38, 40, 47, 53, 60, 63, 64, 71} . Of these nine trials, JiVitA-3 trial ^{53, 54, 73} and Ostadrahimi ⁶⁴ provided supplements to mothers from pregnancy into postpartum, where the other five trials provided supplements to children. PROCOMIDA ³⁷ provided food to the entire family. Ostadrahimi ⁶⁴ enrolled pregnant women from the 20 ^th week of gestational age and were provided daily fish oil supplements (120 mg docosahexaenoic acid and 180 mg eicosapentaenoic acid) or placebo up to 1 month into the postpartum, with their child being followed-up up to six months of age. At the 6-month assessment of this trial, there were no differences found in neither length (mean difference: 0.12, 95% CI: -0.52, 0.76) or head circumference (mean difference: -0.03, 95% CI: -0.38, 0.30) between the fish oil and placebo groups. Mofid ³⁸ found that deworming interventions provided to mothers who tested positive for soil-transmitted helminth infection at baseline had a positive impact on mean length gain (Mean difference: 0.8; 95% CI: 0.1, 1.4) and LAZ (mean difference:0.5; 95% CI: 0.2, 0.8) of infants at six months of age.

Three of the five trials investigated the effectiveness of micronutrient supplements administered directly to children ^{47, 60, 63} . In Feliciano ⁴⁷ , three different dosages of Vitamin D supplements (daily dose of 100, 200, and 400 IU) were provided to Chinese infants from birth up to six months of age; at the 6-months assessment, differences in length between the three groups were observed. Another placebo-controlled trial ⁶³ conducted in Bangladesh found that daily zinc supplements (5 mg) to children between the age of one month to six months did not change the length or head circumference.

There were two trials that explored the role of food supplements to children. Simondon et al. ⁴⁰ was a multi-national trial (Congo, Sengal, Bolivia, and New Caledonia) that randomized four-month old infants to either cereal-based precooked porridge fortified with MMN or the control group consisting of local food. The mean consumption of supplement varied from 133 to 189 kcal/day. There were no differences in length (cm) between the supplemented and control groups in all four countries at six months of age. In Lonnerdal et al. ⁶⁰ , one-month old infants of non-breastfeeding mothers were randomized to receive regular formula or formula fortified with bovine osteopontin (65 or 130 mg/L). There were no differences in length or head circumference between children who were randomized to different formula groups. This trial also recruited infants whose mothers had expressed the desire to exclusively breastfeed up to six months of age and used this breastfeeding group as a non-randomized control. The breastfeeding group had a higher mean head circumference but similar length at six months of age.

All data underlying the results are available as part of the article and no additional source data are required.

Open Science Framework: Interventions to improve linear growth during exclusive breastfeeding life-stage for children aged 0–6 months living in low- and middle-income countries: a systematic review and network and pairwise meta-analyses. https://doi.org/10.17605/OSF.IO/46HMQ ²⁵

File ‘EBF period NMA - Supplementary tables and figures - v2.0’ contains the following extended data:

Open Science Framework: PRISMA checklist for “Interventions to improve linear growth during exclusive breastfeeding life-stage for children aged 0–6 months living in low- and middle-income countries: a systematic review and network and pairwise meta-analyses.” https://doi.org/10.17605/OSF.IO/46HMQ ²⁵ .

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).