We conducted a comprehensive search, including the Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library; MEDLINE via PubMed (1980 to 30 September 2023); and Embase (1980 to 30 September 2023) using each PICO (patient/population, intervention, comparison and outcomes) term. The details of the search strategy have been provided in Supplementary Table S1. We did not apply language restrictions. We searched clinical trial registries for ongoing and recently completed clinical trials.¹

We also searched for abstracts from key nutritional annual meetings. We strived for additional citations by using the references in the articles retrieved through the searches. However, we did not contact subject experts to identify unpublished and ongoing studies (as pre-specified in the protocol), as some of the studies were already in the clinical trial registry.

Two authors independently screened the titles and abstracts of articles identified by searches for eligibility. Data extraction from each included study was carried out using a pre-designed data extraction form. Disagreements were resolved through discussion with a third author.

We used the ‘Risk of bias’ assessment tool and the criteria set out in the Cochrane Handbook for Systematic Reviews of Interventions to assess the risk of bias for included studies (23). We also looked for sources of bias originating from differences between individual RCTs and cluster RCTs (e.g., the relationship between allocation concealment and recruitment bias may be greater in cluster RCTs). Two authors independently assessed the risk of bias in the included studies, and any disagreement was resolved through discussion with a third author.

We assessed reporting biases by trying to identify whether the study was included in a trial registry, whether a protocol is available, and whether the Methods section provides a list of outcomes. We compared the list of outcomes from those sources vs. the outcomes reported in the published article. The inverted funnel was constructed to check for possible publication bias.

We performed statistical analysis according to statistical guidelines referenced in the Cochrane Handbook for Systematic Reviews of Interventions (23). For dichotomous outcomes, we expressed measures of effects as typical risk ratios (RRs) and typical risk differences (RDs) with 95% confidence intervals (CIs). For continuous outcomes, we expressed measures of effect as weighted mean differences (MDs) with 95% CIs. We used the generic inverse variance method in Review Manager 5 to perform a meta-analysis using inflated variances (24). Considering the types of MMN supplements (different preparations and different schedules) as a random factor, a random-effects model was used for all the analyses. We assessed statistical heterogeneity via visual inspection of forest plots of included trials, using the chi-square test and the I² statistic. We used the following cutoffs for the results of the I² test: < 50% low, 50 to 74% moderate, and ≥ 75% high heterogeneity. We attempted to identify the reason for heterogeneity by conducting either subgroup analyses or sensitivity analyses.

We used the GRADE approach, as outlined in the GRADE Handbook, to assess the certainty of evidence (25). We used GRADEproGDT (GRADEpro 2016) to create a ‘Summary of findings’ table to report the certainty of evidence (26).

We conducted subgroup analyses to explore heterogeneity. We analyzed the effects of MMN intervention in the following subgroups.

We conducted sensitivity analyses to assess the impact of a high risk of bias on the outcome of meta-analyses by adding studies with a high risk of bias to pooled studies with a low risk of bias. For completeness of sensitivity analysis, we also employed the leave-out one trial sensitivity analysis method to check the robustness of our meta-analysis for primary outcomes. The steps are as follows: (1) remove the first of the K studies and conduct the meta-analysis on the remaining K-1 studies; (2) remove the second of the K studies and conduct the meta-analysis on the remaining K-1 studies; (3) continue this process until there are K distinct meta-analyses (each with K-1 studies).

If the results of the K meta-analyses in the leave-one-out method are consistent, then there is confidence that the overall meta-analysis is robust.

Of the 11,832 total citations retrieved, the full text of 42 articles was assessed for eligibility, and 30 were excluded for various reasons (Figure 1). The reasons were as follows: preconception supplementation of intervention was not studied (not part of ante-natal supplementation) [n = 24], outcomes of interest not studied [n = 2], MMN not as per standard (UNIMMAP) criteria and outcome of interest not studied [n = 1], active control other than IFA [n = 1], reported outcome at 6 years [n = 1], and duplicate outcome reported [n = 1]. Hence, finally, 12 studies [factorial design = 1] with data from 11,391 participants [intervention = 5,824; control = 5,680] were included (10–21).

The studies were conducted in the following countries: Vietnam, India, Pakistan, Congo, Guatemala, Gambia, and Indonesia. One study was not registered in any clinical trial registry (20). Another study was registered with the Thai Clinical Trial Registry (TCTR), but we could not retrieve the details (10). In three studies, the supplementation started approximately 3 months or so before conception (17–19). In the remaining nine studies, the duration of supplementation was variable before conception (10–16, 20, 21). In all the studies, MMN supplementation was stopped once pregnancy was confirmed, and IFA was continued until delivery. In one study (4 arm) intervention, the age range of participants varied from 16 to 45 years, and all were non-pregnant women. Two studies used lipid-based supplementations as MMN sources (16, 17). Two studies used sachet preparation of MMN (16, 17), two used tablet preparation (11, 21), and the remaining used MMN in capsule form. Table 1 describes other characteristics of the included studies.

The details are provided in the Supplementary Figure S1. Random sequence generation was unclear in one study (20). Five studies were open-label (16–19, 21). Six studies reported higher attrition rates (11–16). Three studies were found to report the outcomes selectively (10, 12, 20). Overall, the included studies were assessed as having a low to moderate risk of bias.

We constructed funnel plots for all three domains (birth weight, length, and head circumference) of the primary outcome (anthropometry at birth) to assess for publication bias. Asymmetry in the funnel plot was noted for the birth weight outcome, for which sensitivity analysis was conducted, but the result did not change (Supplementary Figure S2).

The evidence generated was of “very low certainty” for all the primary outcomes (newborn anthropometric parameters—birth weight, length, and head circumference). For the secondary outcomes, the evidence generated was of “very low certainty” for maternal weight gain and maternal and fetal death; “low certainty” for preterm delivery and neonatal death; and “moderate certainty” for LBW, SGA, and maternal anemia. A detailed analysis of the summary of evidence is provided in Table 2.

After an extensive search of the literature, we found 12 studies eligible for inclusion (data of 11,391 participants). For the primary outcomes, there was no significant difference in the birth weight [MD, 35.61 (95% CI, −7.83 to 79.06), p = 0.11], birth length [MD, 0.19 (95% CI, −0.03 to 0.42), p = 0.09], and head circumference [MD, −0.25 (95% CI, −0.64 to −0.14), p = 0.22] between the MMN and control groups. The GRADE evidence generated was of “very low certainty.” For all the secondary outcomes (except for SGA and LBW), the difference between the MMN and control groups was not significant. There was a significant decrease in the rate of SGA and LBW newborns in the MMN group, and the evidence generated was of “moderate certainty.”

In the subgroup analysis, studies providing lipid-based MMN supplementation in sachet form found significant differences in birth weight [MD, 58.77 g] and length [MD, 0.36 cm] between the MMN and control groups. This is in agreement with previously published data. In the Cochrane review (four trials from LMICs, 8,018 pregnant women) published in 2018, the authors found that newborns in the lipid-based MMN group had a slightly higher mean birth weight (MD 53.28 g) and length (MD 0.24 cm) (27). In a recent trial from Pakistan, the authors included 60 underweight women with pre-eclampsia (28). The lipid-based MMN supplementation group had a significantly higher birth weight (mean difference (MD), 121 g) and length (MD, 0.57 cm) compared to the control group.

There have been conflicting results from trials on MMN vs. IFA supplementation started during the preconception period with some showing benefits, whereas others did not. In one trial from Vietnam, there were no significant differences in the growth and development scores in the offspring between the MMN and IFA groups (14). However, another multi-country three-arm trial found that MMN supplementation led to a higher birth length and less number of SGA newborns with a low rate of stunting (16). However in this trial, for unknown reasons, the Guatemala site did not show a significant difference in the above outcomes between the MMN and IFA groups. In another trial from South Asia, the MMN improved birth weight and reduced stunting and wasting in newborns compared to IFA (18). In this trial, a better effect was seen when the supplementation was initiated >3 months before conception. In the follow-up at 24 months of age, there was no significant difference between MMN and standard care groups (19). The reason for finding no difference might have been due to a small sample size resulting from a high loss of follow-up in the study population. In another large trial, there was no difference in the linear growth and cognitive development between offsprings of MMN (supplemented as capsules) and IFA groups at 2-year follow-up (11). The same group of authors did not find any difference in the postpartum depressive score among mothers in either of the groups (14). A trial assessing placental function in MMN and IFA supplemented found no significant difference, and the same effect was found for the fetal and birth outcomes (15). However, a trial from Vietnam has findings that contrast with the above trials in which simple FA supplementation was in no way different (with regard to the birth weight) from either MMN or IFA supplementation started during the preconception period (10). However, the authors could not provide a suitable explanation for the same. A trial from Indonesia found that fetal survival was significantly better in the MMN-supplemented group (12). A trial from Turkey found no difference between IFA- and MMN-supplemented groups in hemoglobin concentration. In a trial from Indonesia, the fetal length was better in those supplemented with MMNs during the preconception period (20).

In the present systematic review, an overall consistent effect of MMN supplementation was not found. A 17% decrease was observed in the rates of LBW and SGA without affecting other neonatal and maternal outcomes. These findings are similar to those of the findings from antenatal supplementation of MMNs, where the reduction was ≥10% (3, 8, 9). This implies that the differences are more pronounced in the group of children born below the limits considered by the WHO standards as low weight/age and low height/age. However, getting a straight plausible answer for these discrepancies is not an easy task. However, possible explanations, as agreed upon by other researchers, include diversities in the composition and supplementation of MMN formulations or influences from effect modifiers (3, 8, 9). The latter could be the age at enrollment [<20 years vs. ≥20 years, parity (primiparous vs. multiparous)], body mass index (BMI) (<18.5 [underweight] vs. ≥18.5 [not underweight]), maternal height (short stature [<150 cm] vs. normal stature [≥150 cm]), maternal education (none vs. some), and maternal anemia (hemoglobin) status at enrollment (<11 g/dL [anemic] vs. ≥11 g/dL [non-anemic]), gestational age at enrollment (<13 weeks vs. ≥13 weeks), and region (Africa vs. Asia) (8).

In an individual patient data meta-analysis including data from trials on antenatal MMN, the authors observed improved survival for female neonates and greater birth-outcome benefits for infants born to malnourished and anemic pregnant women (9). In addition, early initiation in pregnancy and high adherence to the supplements also provided greater overall benefits.

The studies were variable in many aspects (blinding of participants and outcome assessors, type and dose schedule of the supplements, duration of administration, and outcome measurements). Many were open-label studies and had high attrition rates. Of the 12 included studies, 11 were conducted in the Asian population.

Good-quality trials need to be designed to answer the research questions related to MMN supplementation in diverse settings. Future studies should be more uniform regarding the type of MMN and their supplements (composition, dose, schedule, and duration of supplementation). More follow-up data on neurodevelopmental outcomes in the offspring should be included. Future studies should also include cost–benefit outcomes. More data should come from other LMICs outside the Asian population.