Bulletin of the World Health Organization

Maternal multiple micronutrient supplementation and pregnancy outcomes in developing countries: meta-analysis and meta-regression

Kosuke Kawai a, Donna Spiegelman a, Anuraj H Shankar b & Wafaie W Fawzi b

a. Department of Epidemiology, Harvard School of Public Health, 677 Huntington Ave., Boston, MA 02115, United States of America (USA).
b. Department of Nutrition, Harvard School of Public Health, Boston, USA.

Correspondence to Kosuke Kawai (e-mail: kkawai8@gmail.com).

(Submitted: 19 October 2010 – Revised version received: 23 February 2011 – Accepted: 25 February 2011 – Published online: 21 March 2011.)

Bulletin of the World Health Organization 2011;89:402-411B. doi: 10.2471/BLT.10.083758

Introduction

Every year more than 20 million infants are born with low birth weight worldwide1. About 3.6 million infants die during the neonatal period.2 Two thirds of these deaths occur in southern Asia and sub-Saharan Africa. More than one third of child deaths are thought to be attributable to maternal and child undernutrition.3 Deficiencies in micronutrients such as folate, iron and zinc and vitamins A, B6, B12, C, E and riboflavin are highly prevalent and may occur concurrently among pregnant women.3 Micronutrient deficiencies result from inadequate intake of meat, fruits and vegetables, and infections can also be a cause. Multiple micronutrient supplementation in pregnant women may be a promising strategy for reducing adverse pregnancy outcomes through improved maternal nutritional and immune status.4,5 The World Health Organization (WHO) currently recommends iron and folic acid supplementation to reduce the risk of iron deficiency anaemia among pregnant women. Since many developing countries already have systems in place for the delivery of iron and folic acid supplements, micronutrient supplements could be provided at little additional cost.6

Several systematic reviews of trials examining the effects of maternal multiple micronutrient supplementation have been conducted,710 but they have had limitations. Although some researchers have raised concerns that micronutrient supplementation may increase perinatal mortality, none of the previous review articles has adequately addressed this issue.79,11 None has examined the potential sources of heterogeneity in the effect of supplementation on perinatal mortality. The effects of maternal micronutrient supplementation on perinatal mortality and other pregnancy outcomes can differ depending on trial characteristics and study population. An updated systematic review is essential to provide the basis for future research and for a discussion of policy implications. We conducted a systematic review of trials comparing the effect of maternal multiple micronutrient supplementation with that of iron and folic acid supplementation on pregnancy outcomes in developing countries. We also conducted subgroup meta-analysis and meta-regression to explore sources of heterogeneity.

Methods

Search strategy and inclusion criteria

We searched MEDLINE and EMBASE up to 1 August 2010 and identified potentially relevant published trials using the combination of medical subject headings (MeSH) and text words denoting micronutrient supplements and pregnancy outcomes. We used the terms micronutrient, multivitamin, vitamin, mineral and supplement in combination with pregnancy, birth, newborn, infant, low birth weight, preterm, fetal growth, small-for-gestational-age, perinatal mortality and neonatal mortality. We also searched the references cited by the retrieved articles for additional references.

We applied the following inclusion criteria: (i) only randomized controlled trials conducted in developing countries; (ii) only trials that compared an intervention group receiving multiple micronutrient supplements (defined as more than three micronutrients) with a control group receiving iron and folic acid, and (iii) only trials that examined any of the following outcomes: birth weight, low birth weight (birth weight < 2500 g), preterm birth (birth before the 37th week of gestation), small size for gestational age (birth weight below the 10th percentile of weight for gestational age), perinatal death (death from the 28th week of gestation through the first week after delivery) and neonatal death (death within 28 days of delivery).

We included only trials in developing countries because micronutrient malnutrition among pregnant women is much more common in developing countries than in industrialized countries. Furthermore, no recommendations for prenatal micronutrient supplements exist in most developing countries, whereas such supplements are routinely recommended in some developed countries.12

We defined a “multiple micronutrient supplement” as a single tablet containing more than three different micronutrients; micronutrient-fortified powders, foods and beverages were not included in the definition.

Data extraction and quality assessment

We extracted information on the characteristics of the trials and their participants, interventions, outcome measures and methodological quality. Our criteria for assessing methodological quality included randomization, allocation concealment, blinding, completeness of follow-up and compliance with study regimens. We also extracted the results of randomized controlled trials included in previous reviews.

Statistical analysis

We applied random effects models to estimate the pooled relative risks (RRs) or mean differences and their respective 95% confidence intervals (CIs).13 The natural logarithms of the RRs and their corresponding standard errors from individual trials were used to compute the pooled estimates. We employed the I2 statistic to determine the proportion of the total variation among studies due to heterogeneity rather than chance.14 To evaluate possible publication bias, we visually inspected the funnel plot for asymmetry and performed the Begg’s rank correlation test and Egger’s linear regression test.

To investigate the sources of heterogeneity, we conducted subgroup meta-analysis and meta-regression analysis.15 We examined the effects of micronutrient supplementation according to the following dichotomized characteristics: prevalence of maternal underweight (body mass index [BMI, or weight in kg divided by the square of the height in metres] < 18.5), < 20% or ≥ 20%; average height, < 155 cm or ≥ 155 cm; proportion of primiparous women, < 35% or ≥ 35%) and proportion of women without formal education, < 50% or ≥ 50%. Additionally, we examined the effect modification of different iron doses as a function of treatment group and average gestational age at initiation of supplementation (< 20 weeks or ≥ 20 weeks). For perinatal mortality, we also examined the effect of micronutrient supplementation on the risk of large size for gestational age (RR < 1.10 or ≥ 1.10) and birth weight (mean difference in birth weight by treatment group, < 50 g or ≥ 50 g). We performed random effects meta-regression using weights proportional to the inverse variance. We employed residual I2 to estimate residual heterogeneity between studies after adjusting for the covariate of interest in the meta-regression. All statistical analyses were performed using STATA version 11 (StataCorp. LP, College Station, United States of America).

Results

We identified 27 potentially relevant articles for detailed review (Fig. 1). We excluded 10 articles because they were conducted in developed countries, used supplements containing fewer than three micronutrients or did not assess outcomes of interest. The remaining 17 articles provided data for meta-analysis.9,10,1631 Three of them were systematic reviews9,10,31 from which we extracted the results of trials in Indonesia (Indramayu) and Bangladesh.9,10

Fig. 1. Selection of studies included in systematic review of randomized controlled trials on maternal multiple micronutrient supplementation and pregnancy outcomes in developing countries

Table 1 shows the characteristics of the 15 trials (n = 64 244) that compared the effect of micronutrient supplementation with that of iron and folic acid supplementation on pregnancy outcomes. All trials were double-blinded and showed high compliance with the study regimens. While loss to follow-up was minimal in most trials, in Mexico, Niger and Zimbabwe more than 20% of trial participants were lost to follow-up. A hospital-based trial in India was restricted to only undernourished women and two trials conducted in the United Republic of Tanzania were restricted to either HIV-positive or HIV-negative women. Most of the trials used the international multiple micronutrient preparation (UNIMMAP) of the United Nations Children’s Fund/World Health Organization/United Nations University, which contains the recommended dietary allowance (RDA) of 15 vitamins and minerals. Because vitamins can enhance iron absorption and utilization, the UNIMMAP contains 30 mg of iron, or half the dose of iron in the iron and folic acid supplements. The trials conducted in Mexico, Nepal (Sarlahi district) and Zimbabwe used the RDA of multiple micronutrients, like the UNIMMAP, whereas two trials conducted in the United Republic of Tanzania used multivitamin supplements containing multiples of the RDA. In the trial conducted in India, supplements contained the RDA of 29 micronutrients.

Pooled analyses of 15 trials showed that maternal micronutrient supplementation increased birth weight (pooled mean difference: 44 g; 95% CI: 28–60; Table 2). Pregnant women who received micronutrient supplements were less likely to deliver low-birth-weight infants (pooled RR: 0.86, 95% CI: 0.79–0.93) or small-for-gestational-age infants (pooled RR: 0.85; 95% CI: 0.78–0.93) than women who received iron and folic acid supplements. Micronutrient supplementation had no effect on preterm delivery (pooled RR: 0.99; 95% CI: 0.95–1.03). We examined sources of heterogeneity for effects on birth weight, low birth weight and small size for gestational age (Table 3, Table 4 and Table 5, available at: http://www.who.int/bulletin/volumes/89/6/10-083758). None of these three outcomes appeared to be modified by maternal education, underweight, height, parity, the average timing of initiation of supplements or iron dosage (P for test for heterogeneity > 0.05). However, trials in which micronutrient supplementation was initiated after an average gestational age of 20 weeks appeared to show a greater beneficial effect of supplementation with respect to low birth weight (RR: 0.76; 95% CI: 0.64–0.89) and small size for gestational age (RR: 0.77; 95% CI: 0.65–0.91), although the test for heterogeneity did not yield statistical significance.

The effect of micronutrient supplementation on perinatal and neonatal mortality was assessed in 11 randomized controlled trials (Table 6). A pooled estimate showed no overall effect of micronutrient supplementation on perinatal mortality (RR: 1.05; 95% CI: 0.90–1.22) or neonatal mortality (RR: 1.08; 95% CI: 0.92–1.26; Fig. 2). However, substantial heterogeneity of effect estimates on perinatal mortality was evident (I2 = 58%; 95% CI: 19–79; P for heterogeneity = 0.008). Subgroup and meta-regression analyses showed that differences between trials in maternal underweight, maternal height, parity and iron dosage did not explain the heterogeneity of effect estimates on perinatal mortality (Table 7). Furthermore, differences in the effect of supplementation on the risk of large size for gestational age did not explain the heterogeneity. In univariate meta-regression analysis, adverse effects on perinatal mortality were found in trials conducted in settings where most mothers had no formal education (P = 0.009). All three trials conducted in populations in which more than 50% of mothers had no formal education (Burkina Faso, Sarlahi in Nepal and Pakistan) showed adverse effects on perinatal mortality (pooled RR: 1.41; 95% CI: 1.14–1.75). The average gestational age at the initiation of supplementation may have contributed to the heterogeneity in the effect estimates, and early initiation was associated with harmful effects (univariate meta-regression, P = 0.003). The pooled RR was 1.28 (95% CI: 1.10–1.49) for trials in which supplementation was on average initiated before the 20th weeks of gestation and 0.88 (95% CI: 0.80, 0.97) for trials in which supplementation was initiated after the 20th week. Maternal education and average gestational age at supplement initiation were correlated; trials conducted among mothers with no education were also those that initiated supplementation early. When we constructed a multivariate meta-regression model including maternal education and gestational age at initiation of supplementation, both covariates became non-significant (timing of initiation, P = 0.09; maternal education, P = 0.30; residual I2 = 0%; residual P for heterogeneity = 0.51). We also conducted meta-regression analysis to explore the effect of micronutrient supplementation on neonatal mortality. Similarly, we found that maternal educational level (univariate meta-regression; P = 0.01) or average gestational age at initiation of supplementation (P = 0.02) may have contributed to the heterogeneous effect on neonatal mortality.

Fig. 2. Effect of maternal micronutrient supplementation on perinatal and neonatal mortality in randomized controlled trials in developing countries

CI, confidence interval; HIV, human immunodeficiency virus; RR, relative risk.

Neither the Egger test nor the Begg test for publication bias showed statistical significance (P > 0.05 for both tests) for an effect on low birth weight, birth weight, small size for gestational age, preterm delivery and perinatal mortality. In the case of neonatal mortality, we found some evidence of publication bias based on the Egger test (P = 0.02) but not on the Begg test (P = 0.28).

Discussion

Our study is consistent with recent systematic reviews in showing that maternal micronutrient supplementation can reduce the risk of having an infant with low birth weight.8,31 While micronutrient supplementation showed no overall effect on perinatal mortality, several trials reported non-significant adverse effects. We found that maternal educational level or gestational age at initiation of supplementation may have contributed to the heterogeneous effects on perinatal mortality.

Several biological mechanisms can explain the beneficial effects of micronutrient supplementation on fetal growth. Women require more vitamins and minerals during pregnancy and supplements can improve their nutritional and haemoglobin status. Supplements also help improve and maintain functional immunity.32 Deficiencies of B-complex vitamins and folate are prevalent and may be a major cause of homocysteinaemia.4 Elevated homocysteine levels can lead to endothelial cell dysfunction and affect placental function. Thus, micronutrient supplements can help maintain normal homocysteine levels. Many vitamins and minerals also play important roles in gene regulation as well as in cellular metabolism and fetal growth.33

Why does micronutrient supplementation appear to be associated with increased risk of perinatal death in some trials? Christian et al. have hypothesized that such adverse effects may be due to an increased risk of birth asphyxia or of cephalopelvic disproportion among infants who are large for gestational age.18 However, in our study the heterogeneity in the effect on perinatal mortality was not explained by the prevalence of maternal underweight, average maternal height or the percentage of primiparous women. In a trial in Burkina Faso, half of the perinatal deaths were due to prematurity. This led the researchers to conclude that an increase in perinatal mortality cannot be entirely due to the complications associated with delivering an infant too large for gestational age.28 Furthermore, large head size can increase the risk of obstructed delivery, but a previous review found no effect of micronutrient supplementation on neonatal head circumference.10 Alternatively, the effects on perinatal mortality could stem from the use of a lower dose of iron in the micronutrient group than in the iron and folic acid group.27 However, differences in the dose of iron did not explain heterogeneity in our analysis.

We found that all trials that reported an adverse effect on perinatal mortality were conducted in poor rural settings where most mothers had no education. Maternal educational level is likely to be a proxy for unmeasured characteristics. Low maternal education may be a correlate of a greater likelihood of delivering at home, limited access to health facilities, limited availability of skilled birth attendants and maternal and newborn care of lesser quality.34 Mothers with limited education are less likely to seek neonatal care and more likely to be at risk of adverse pregnancy outcomes. Furthermore, in most developing countries access to quality perinatal health care differs substantially between rural and urban areas.35 Our findings suggest that micronutrient supplements may need to be delivered under adequate obstetric and postnatal care. Large cluster randomized trials with a stepped wedge design may need to be conducted in rural settings to assess the safety and efficacy of micronutrient supplementation in the context of programmes for improving obstetric and postnatal care. In many parts of the developing world, enhancing access to obstetric care, improving postnatal care and empowering mothers through community health workers are essential measures for reducing perinatal and neonatal mortality.36,37

There is no clear explanation for the association between multiple micronutrient supplementation and a higher risk of perinatal mortality found in trials in which supplementation began in the first and early second trimester of pregnancy. This finding needs to be interpreted with caution because the meta-regression explored the potential relationship between treatment effect and the timing of supplement initiation (e.g. average gestational week) across trials. Thus, we are unable to assess whether a relationship exists or not between treatment effect and the timing of supplement initiation within trials.15 It is possible that supplementation for a longer period increased the risk of complications during labour by increasing infants’ size for gestational age. Alternatively, early supplementation could have altered metabolic regulation and led to complications during pregnancy and consequently to perinatal death. Another possibility is that micronutrient supplementation prevented early spontaneous abortion and allowed mothers to carry frail fetuses to much later stages of pregnancy, with a resulting spurious increase in the number of perinatal deaths. Minerals can interact with desoxyribonucleic acid (DNA) and disrupt ligand binding or protein function, or induce oxidative damage on embryotic tissue.38 There is clearly a need to elucidate the biological mechanisms behind the potentially detrimental effects of micronutrient supplementation.

Our study has several limitations. Factors that we did not examine, such as the prevalence of maternal infections, including malaria and hookworm, and dietary nutrient intake, may have accounted for the heterogeneity of effect estimates. As shown by a trial in Nepal, zinc can reduce the beneficial effects of iron supplements through biochemical interactions that are possibly influenced by whether or not supplements are taken with a meal.18,39,40 The heterogeneity of effect estimates could also depend on differences in the quality of the trials; however, most trials showed good compliance with the study regimen and high participant retention. We were unable to verify the composition and dosage of the micronutrient supplements. Most trials used the RDA of each micronutrient, an amount considered sufficient to meet the requirements of most healthy individuals in industrialized countries. In developing countries pregnant women may require higher doses because of their poorer nutritional status and higher rates of infection. However, a recent trial found that multivitamin supplements containing the RDA of each component may be as effective as those containing multiple doses of the RDA in reducing the risk of adverse pregnancy outcomes among HIV-infected women.41

There is insufficient evidence to recommend routine prenatal multiple micronutrient supplementation for women in developing countries. Although our study provides valuable insights into the heterogeneous effects of micronutrient supplements on perinatal mortality, more research is needed to explain why some trials found multiple micronutrient supplementation to be associated with higher perinatal mortality. Few studies have examined the effects of micronutrient supplementation on long-term child health outcomes, such as child mortality, morbidity, growth and cognitive development. A large trial in Indonesia showed that prenatal micronutrient supplementation was associated with a significant 18% reduction in early infant mortality.26 Examining long-term outcomes is important because greater infant survival or other beneficial effects may not be reflected in birth weight. Micronutrient supplementation before pregnancy also warrants further research.42,43 Among Bhutanese refugees dependent on food aid, the incidence of low birth weight declined from 16% to 8% after 2 to 3 years of implementation of micronutrient-fortified blended foods.44

In conclusion, we found that maternal micronutrient supplementation can reduce the risk of low birth weight but has no overall effect on perinatal mortality in developing countries. Non-significant detrimental effects on perinatal mortality were reported in some trials conducted in poor rural settings. This suggests that in such settings supplements may need to be delivered in the context of programmes for improving obstetric and postnatal care. More research is needed to address the safety, efficacy and effective delivery of maternal micronutrient supplementation.


Competing interests:

None declared.

References

Share