Interventions to improve birth outcomes of pregnant women living in low- and middle-income countries: a systematic review and network meta-analysis

Introduction

Despite global substantial progresses that have been made towards improving maternal, newborn, and child health (MNCH) in the last two decades, adverse birth outcomes such as preterm birth and low birthweight still remain as an important global health challenge, particularly in low- and middle-income countries (LMIC)^1–3. Determinants of these challenges are multifaceted^4–7. Pregnant women in LMICs have a higher risk of nutritional deficiencies, stemming from physiological changes that involve fetal development and growth resulting in an increased demand for nutrients^4,5. Poor water, sanitation, and hygiene (WASH) control can also increase likelihood for infectious diseases, including intestinal worm infections that may contribute to conditions, such as anemia, which negatively affects fetal survival and growth^6,8,9. Poor maternal health during pregnancy is associated with preterm birth (<37 gestation weeks) and low birthweight (<2500 g), and these adverse birth outcomes are associated with adverse neonatal events, such as respiratory distress syndrome, neurocognitive impairment, poor linear growth (stunting), and overall mortality^1,2,10,11.

Several reviews have aimed to assess the effectiveness of various promising interventions for pregnant women (Table 1). Despite the extensive research conducted to date, the comparative effectiveness of interventions remains unclear across different domains, such as micronutrients, balanced energy protein supplements, maternal education, deworming, and WASH. Few clinical trials have directly compared interventions across domains. Rather, the majority of clinical trials has only compared interventions within a domain. Similarly, most summaries of the evidence for pregnancy interventions have used traditional pairwise meta-analysis, allowing only for the quantitative assessment of a single intervention versus a comparator. Thus far, no attempts have been made to synthesize the evidence indirectly in order to make quantitative comparison of interventions that have not been directly compared in studies.

Recognizing the paucity of direct head-to-head randomized clinical trials (RCTs) between existing interventions, a network meta-analysis can be used to summarize the entirety of evidence for pregnancy interventions. A network of interventions connected via the comparisons that have been made in head-to-head trials can be constructed, and where there is a path from one intervention to another, these interventions can be compared indirectly via some common comparators^12–16. In addition, where both direct and indirect evidence exists, the indirect evidence can be used to strengthen the inferences for the particular comparison. This is particularly important for pregnancy interventions because many head-to-head trials of active interventions have limited sample sizes. Furthermore, network meta-analysis allows us to simultaneously analyze all potential treatment options and make full use of the available evidence within a single analysis.

The purpose of this study was to assess the comparative effectiveness across intervention domains in micronutrient supplements, balanced energy protein supplements, deworming, maternal education, and WASH interventions using network meta-analysis. Effectiveness of interventions are determined by the following outcome indicators: preterm birth, low birthweight, and birthweight for LMIC-based pregnant women.

Methods

Our analysis and report was designed and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) extension to network meta-analysis¹⁷. The protocol for this study is registered on PROSPERO (CRD42018110446).

Results

We identified 5,297 abstracts from our database searches and hand searches of the bibliography of the published reviews (Figure 1). Of these, 377 studies underwent a full-text review, and 132 papers reporting on 101 trials met our inclusion criteria. In total, these trials included 206,531 pregnant women that were randomized to 245 unique interventions (Figure 2). The list of included and excluded studies (Extended data, Supplementary Tables 4 and 5)¹⁹, as well as the trial and patient characteristics of the included studies (Extended data, Supplementary Tables 6 and 7)¹⁹ are provided in the Extended data. Geographically, most trials were conducted in South Eastern Asian (n = 38 trials) and African (n = 26 trials) countries, with individual randomization (i.e. non-cluster trials, n = 85 trials) and double blinding (n = 52 trials) being the most common methodological features. Micronutrient supplements was the most common intervention domain that was investigated (n = 79 trials); only a few of these micronutrient trials compared interventions from other domains, such as balanced energy protein supplements (n = 15 trials) and deworming (n = 6 trials). Maternal education was captured in a few trials (n=3) where the education component was in the form of counseling, lifestyle education, and participatory learning action (PLA) with government-mandated women’s groups. PLA involved awareness of the problem of LBW and malnutrition, and strategies to overcome barriers to improved health and nutrition. There were no WASH trials reporting on the analyzed birth outcomes.

Figure 1. Study selection.

Figure 2. Overall network of the comparisons between interventions for pregnancy.

Each node (circle) represents an intervention, each line represents a direct comparison between interventions, with the lines with width representing the number of trials with the direct comparisons in question (i.e. thicker width represents a direct comparison with larger numbers of trials). The different intervention domains are indicated with the following colors: blue for micronutrient supplements; brown for balanced energy protein supplements; yellow for education and counseling interventions; and green for deworming interventions. Vit, Vitamin; IFA, iron and folic Acid; LNS, lipid-based nutrient supplements; Fort, fortification; MMN, multiple micronutrients.

In most trials, interventions were provided to pregnant women from enrollment until delivery (n = 87 trials). These trials generally involved women who were in the later part of their gestational age. For instance, only 5 trials enrolled women from or before conception (Owens⁴⁶, The women First Trial⁴⁷, CAP Trial⁴⁸, PRECONCEPT⁴⁹, and Brabin⁵⁰), while the majority of trials recruited women who were in the later trimesters, such as the 2^nd and 3^rd trimesters (n = 69 trials).

Preterm birth (<37 weeks of gestational age)

The preterm birth network (Extended data, Supplementary Figure 1)¹⁹ included 64 trials consisting of 85,546 pregnant women randomized to 152 intervention arms (ten cluster trials consisting of 1,998 clusters and 20,218 pregnant women). From the primary analysis, that included both cluster and non-cluster randomized clinical trials, only few interventions showed superiority over standard-of-care for preterm birth (Figure 3). For instance, compared to standard-of-care, intake of 1500 kcal of local food per day showed an RR of 0.36 (95% CrI: 0.16, 0.77) and calcium showed an RR of 0.76 (95% CrI: 0.56, 0.99). Other micronutrient supplements such as folic acid (RR: 0.71, 95% CrI: 0.43, 1.09), iron (RR: 0.70, 95% CrI: 0.47, 1.01), zinc (RR: 0.67, 95% CrI: 0.41, 1.04), and multiple micronutrients (MMN) (RR: 0.70, 95% CrI: 0.45, 1.02) showed a trend towards lower preterm birth risks compared to standard-of-care, but their Crls overlapped the null effect of 1.00. In comparison to standard-of-care, no balanced energy food supplements, other than 1500 kcal of local food showed reduction in preterm birth risks, and neither did maternal education interventions (e.g. participatory learning action⁵¹).

Figure 3. Forest plot for the effects of interventions on preterm birth, risk ratio.

Vit, Vitamin; IFA, iron and folic Acid; LNS, lipid-based nutrient supplements; Fort, fortification; MMN, multiple micronutrients.

Mean birthweight (kg)

The mean birthweight network (Extended data, Supplementary Figure 3)¹⁹ included of 81 trials that consisted of 130,315 pregnant women randomized to 196 intervention arms. Of these 81 trials, 14 were cluster trials that randomized 1,354 clusters (57,483 pregnant women) to 35 intervention arms. The results of the network meta-analysis on mean birthweight can be found in Figure 4. Among the micronutrient supplementation domain, compared to standard-of-care, MMN (mean difference: 0.27 kg; 95% CrI: 0.09, 0.45 kg), folic acid (mean difference: 0.21 kg; 95% CrI: 0.00, 0.42 kg), iron (mean difference: 0.18 kg; 95% CrI: 0.02, 0.34 kg), and iron + folic acid (IFA) (mean difference: 0.18 kg; 95% CrI: 0.00, 0.36 kg) showed improvements in birthweight. Among the balanced energy food supplements, unfortified lipid-based nutrient supplements of 20 grams (LNS20) showed improvements in birthweight compared to standard-of-care (mean difference: 0.27 kg; 95% CrI: 0.03, 0.51 kg). Deworming and maternal education interventions did not improve mean birth weight; for instance, in comparison to standard-of-care, a single dose of deworming showed a mean difference of 0.02 kg (95% CrI: -0.16, 0.19 kg) and maternal education showed a mean difference of 0.07 kg (95% CrI: -0.27, 0.40 kg).

Figure 4. Forest plot for the effects of interventions on birthweight, mean difference in kg.

Vit, Vitamin; IFA, iron and folic Acid; LNS, lipid-based nutrient supplements; Fort, fortification; MMN multiple micronutrients.

Low birthweight (<2.5 kg)

The low birthweight network (Extended data, Supplementary Figure 5)¹⁹ consisted of 67 trials, with 84,675 patients randomized to 160 intervention arms (eleven cluster trials consisting of 792 clusters and 9,512 pregnant women). The results on low birthweight (kg) outcome can be found in Figure 5. High caloric local food intervention (1500 kcal per day) reduced the risk of low birthweight (RR: 0.17; 95% CrI: 0.09; 0.29). There was a trend towards reduced risks of low birthweight for other interventions such as calcium (RR: 0.87; 95% CrI: 0.69; 1.07), MMN (RR: 0.73; 95% CrI: 0.49, 1.07), and LNS20 (RR: 0.65; 95% CrI: 0.39, 1.07), fortified LNS20 (RR: 0.72, 95% CrI: 0.48, 1.03), but their 95% CrI contained the null effect of 1.00. A single dose of deworming during pregnancy did not show reduction in low birthweight when compared to standard-of-care (RR: 1.15, 95% CrI: 0.83, 1.58).

Figure 5. Forest plot for the effects of interventions on low birthweight, risk ratio.

Vit, Vitamin; IFA, iron and folic Acid; LNS, lipid-based nutrient supplements; Fort, fortification; MMN multiple micronutrients.

Discussion

In this study, we used network meta-analysis to compare the effectiveness of interventions across several domains consisting of micronutrient supplements, balanced energy protein supplements, deworming, maternal education, and WASH interventions that can be provided to pregnant women living in LMICs. Several micronutrient supplements demonstrated decreased risks for preterm birth and/or improve mean birthweight, compared with standard-of-care for pregnant women. For example, MMN interventions showed reduction in preterm birth risks and improved mean birthweight. In comparison to standard-of-care, IFA, calcium, iron, and zinc also demonstrated a trend towards decreasing preterm birth risks. However, the evidence for other intervention domains were limited. For instance, among balanced energy protein supplements, only consumption of 1500 kcal of local food supplement lowered the risks of preterm birth and low birthweight; and only unfortified LNS 20 demonstrated improvement in mean birthweight. Nevertheless, these findings pertaining to balanced energy protein supplements corresponded to only three trials in the study^52–54. There was a limited number of trials available for maternal education and deworming intervention; no WASH trials reporting on preterm birth and birthweight outcomes were available.

The main strength of this study was the use of network meta-analysis to assess the effectiveness of different interventions from a large network of evidence compared to standard-of-care. Unlike previous reviews that have focused on one intervention within a single domain, we used a broad evidence base that included multiple interventions from different domains. As well, appropriate statistical adjustments were made for clustering effects of cluster randomized clinical trials to enable the convergence of cluster and non-cluster trials for our network meta-analysis. Nevertheless, the existing evidence base limited our analyses. Few trials reported low birthweight, and the majority of randomized clinical trial evidence base was confined to a single domain of micronutrient supplementation. Another possible limitation was that there was notable variation in the enrollment of pregnant women in terms of trimesters and gestational age. While we did not find that time of enrollment relative to gestational age was a treatment effect modifier in our analyses, we acknowledge that this variation may have introduced heterogeneity in our meta-analyses. Prior evidence has also demonstrated mixed evidence as to whether the time at which treatment is initiated influences overall treatment efficacy, and this varies by treatment type^25,31,37. Lastly, our assumption of a conservative ICC (0.05) may also have affected the results. However, this was necessary in order to assess for the entire evidence base of interventions for pregnancy, as most cluster randomized trials did not report ICC for each outcome.

Despite these limitations, the findings of this study were generally similar to that of other existing reviews. For instance, among the micronutrient supplements, other reviews have shown that iron (RR=0.82, 95%Crl 0.72, 0.94)⁵⁵ and MMN (RR=0.88, 95%Crl 0.85, 0.90)³¹ reduced the risks of low birthweight versus standard-of-care. Moreover, calcium (RR=0.76, 95%Crl 0.60, 0.97)²⁵ and zinc (RR=0.86, 95%Crl 0.76, 0.97)³⁷ supplements reduced the risks of preterm birth, and we have found that intake of combined MMN reduced the risks of preterm birth and improved mean birthweight. Similar to this study, Salam²⁶ found no improvements in low birthweight and preterm birth for deworming versus standard-of-care. There were no reviews on WASH available that looked at the role of WASH interventions on birthweight and preterm birth.

Our findings identified several directions for future research. First, there is a need to combine interventions that consist of compelling and evidence-based interventions of different domains as a package, moving away from a reductionist approach that is reflected in the majority of clinical trials conducted so far. Instead of a singling out interventions from one domain, there is a need for more evidence of packaged interventions because a combined set of interventions will likely result in the greatest improvement for adverse birth outcomes. Second, more research is needed to assess the longevity of interventions and its effectiveness across multiple life stages. For instance, only 17 out of 101 trials conducted follow-ups of women after birth delivery into the post-partum period. It is also important to note that the median follow-up of pregnant women beyond delivery was 8 weeks and only three trials^23,56,57 conducted follow-ups with women and their newborns up to 6 months of age.

Overall, we identified a number of interventions for pregnancy with clear and compelling supportive evidence for effectiveness for preventing adverse birth outcomes. In midst of the World Health Organization’s Global Nutrition Targets 2025⁵⁸, which focuses on improving maternal, infant, and young children nutrition, national and local MNCH programs should consider adopting and adapting effective interventions identified in this review based on their local resource availability and program priorities. This may provide an opportunity to evaluate the benefits of these interventions in routine practice for pregnancy, and a step towards reaching the 2025 Global Nutrition Target of reducing the global prevalence of low birthweight by 30%⁵⁹.

Data availability