ABSTRACT
Background
Maternal micronutrient supplementation in pregnancy (MMS) has been shown to improve birth weight among infants in low- and middle-income countries. Recent evidence suggests that the survival benefits of MMS are greater for female infants compared to male infants, but the mechanisms leading to differential effects remain unclear.
Objective
The objective of this study was to examine the potential mechanisms through which MMS acts on infant mortality among Tanzanian infants.
Methods
We used data collected from pregnant women and newborns in a randomized, double-blind, placebo-controlled trial of MMS conducted in Tanzania to examine mediators of the effect of MMS on 6-wk infant mortality (NCT00197548). Causal mediation analyses with the counterfactual approach were conducted to assess the contributions of MMS on survival via their effects on birth weight, gestational age, weight-for-gestational age, and the joint effect of gestational age and weight-for-gestational age. The weighting method allowed for interaction between gestational age and weight-for-gestational age.
Results
Among 7486 newborns, the effect of MMS on 6-wk survival was fully mediated (100%) through the joint effect of gestational age and weight-for-gestational age. MMS was also found to have a significant natural indirect effect through increased birth weight (P-value < 0.001) that explained 75% of the total effect on 6-wk mortality. When analyses were stratified by sex, changes in gestational age and weight-for-gestational age fully mediated the mortality effect among female infants (n = 3570), but these mediators only explained 34% of the effect among males (n = 3833).
Conclusions
The potential sex-specific effects of MMS on mortality may be a result of differences in mechanisms related to birth outcomes. In the context of the Tanzanian trial, the observed effect of MMS on 6-wk mortality for female infants was entirely mediated by increased gestation duration and improved intrauterine growth, while these mechanisms did not appear to be major contributors among male infants.
Keywords: maternal nutrition, multiple micronutrient supplementation, infant mortality, mediation, birth outcomes
Introduction
Micronutrient deficiencies are common among pregnant women in low- and middle-income countries (LMICs) caused by inadequate dietary intake and limited dietary diversity (1). In 2016, an estimated 20% of pregnant women globally had iron deficiency anemia, and 15% of pregnant women had vitamin A deficiency (2). Deficiencies of zinc, vitamin B-12, vitamin E, and folate are also suspected to be highly prevalent among pregnant women (2).
Decades of research have shown that micronutrient deficiencies in pregnancy can lead to poor fetal growth and child health outcomes (3). Therefore, antenatal multiple micronutrient supplementation (MMS) has been examined as a potential intervention to improve birth outcomes and child survival. A 2019 Cochrane review determined that across 20 randomized trials, MMS improved birth weight and reduced the risk of small-for-gestational-age (SGA) births (4). In addition, a recent individual patient data meta-analysis found that MMS provided greater survival benefits for female infants compared to males, and MMS had greater beneficial effects on birth outcomes among undernourished and anemic pregnant women (5). The mechanisms through which antenatal MMS affects infant mortality, potentially in a sex-specific manner, are not clear. For example, some have suggested that MMS increases birth size and thus reduces infant mortality; however, it has also been suggested that this increase in birth size may lead to an increased risk of birth complications (6, 7). Thus, understanding the mechanisms of action though which MMS operates may better inform global policy.
To address this evidence gap, we conducted a mediation analysis of the effect of MMS on mortality with data from a randomized trial conducted in Tanzania (8). We hypothesized that the effect of MMS on mortality was mediated by increased gestation duration and improved intrauterine growth. We used a counterfactual approach to mediation, which provides a less biased estimate than traditional product or difference methods (9). This novel method also allows for interaction between mediators, which in this case allowed us to model the interaction between gestational age and size-for-gestational age on mortality. We calculated the natural indirect effect (NIE) of MMS on mortality mediated through birth weight and mediated through the joint action of weight-for-gestational age and gestational age at birth. We also investigated these mediators stratified by sex to examine potential mechanisms leading to sex-differential survival effects of MMS.
Methods
Study design
This analysis used data from a randomized, double-blind trial conducted among HIV-negative pregnant women in Dar es Salaam, Tanzania (ClinicalTrials.gov number, NCT00197548) (8). Briefly, pregnant women were eligible for the trial if they tested negative for HIV infection, planned to stay in the city until 1 y postpartum, and were 12–27 wk pregnant according to self-reported date of last menstrual period (LMP). All participants provided written consent for participation. Study participants were randomly assigned to receive either MMS (including iron-folic acid) or iron-folic acid (IFA) supplements alone. The MMS contained the following nutrients and dosages: 60 mg of iron, 20 mg of thiamin, 20 mg of riboflavin, 25 mg of vitamin B-6, 100 mg of niacin, 50 μg of vitamin B-12, 500 mg of vitamin C, 30 mg of vitamin E, and 0.8 mg of folic acid in tablet form. The IFA arm received 60 mg of iron and 0.8 mg folic acid. The original sample size for the trial was 6000 pregnancies; however, because of low event rates the target sample size was increased to 8468. The trial was approved by the institutional review boards at Muhimbili University of Health and Allied Sciences in Dar es Salaam, the National Institute of Medical Research, and the Harvard TH Chan School of Public Health (8).
At randomization, research staff measured maternal height, weight, and hemoglobin, and completed a questionnaire with all participants that included sociodemographic characteristics. After this baseline visit, monthly clinic visits were conducted until delivery. Research midwives attended the deliveries of the study participants and measured birth weight of the infants to the nearest 10 g. Women who missed monthly clinic appointments were visited at home by research staff. Women and their infants were followed up until 6 wk postpartum.
Statistical analysis
The primary statistical analysis was restricted to singleton live births. We explored four possible mediating pathways for the effect of MMS on 6-wk infant mortality: birth weight, gestational age alone, weight-for-gestational age alone, and the joint mediation of gestational age and weight-for-gestational age. Figure 1 presents directed acyclic graphs for each potential pathway. Gestational age was calculated with use of the self-reported date of LMP. For gestational ages > 300 d, a 300-d gestation was assumed. Preterm birth was defined as those born with a gestational age at birth <37 wk. Size-for-gestational age was calculated with z scores as calculated through use of the INTERGROWTH-21(st) standards (10). SGA births were defined as those below the 10th percentile of birth weight for their gestational age.
FIGURE 1.
Mediators investigated in this analysis. The mediated pathways were investigated in four separate analyses: (A) the effect of multiple micronutrient supplements on 6-wk infant mortality via the mediated path through birth weight; (B) the effect of multiple micronutrients on 6-wk infant mortality via the mediated path through gestational age; (C) the effect of multiple micronutrients on 6-wk infant mortality via the mediated path through weight-for-gestational age; and (D) the effect of multiple micronutrients on 6-wk infant mortality via two parallel, independent, mediated pathways of weight-for-gestational age and gestational age. All mediators in these analyses are estimated as curvilinear terms.
A counterfactual approach to mediation analysis was used to determine the NIE through each mediator, natural direct effect (NDE) through any pathway outside of the mediator, and total effect of MMS on 6-wk infant mortality (9). All mediation analyses were conducted on the additive scale and the effects can be interpreted as risk differences between counterfactuals. For each participant, we assume that a mediator, for example birth weight, could naturally take on different values depending on whether they receive MMS or placebo. The NIE is the change in average mortality had we fixed every infant's exposure to receiving MMS, but artificially manipulated birth weight to take on its natural value under exposure to MMS compared to its natural value under exposure to the placebo. Thus, the NIE is the amount of the total effect of MMS on mortality that can be explained by the mediator.
The NDE compares how the average mortality would change comparing receiving MMS to receiving placebo; if we artificially fix birth weight for each infant to the natural value that we would expect had they received the placebo. Thus, the NDE explains the amount of the effect of MMS on mortality that is not explained by the mediator. We can then calculate the proportion of the effect mediated by birth weight as the fraction of the total effect explained by the indirect effect. The proportion mediated was described as “fully explained” if the NIE accounted for the full magnitude (100%) of the total effect assuming that all confounders between the mediators and outcomes were included in the measured variables.
These mediation analyses assume no modification of the effect of MMS on 6-wk infant mortality by the mediators, which we confirmed empirically using log-binomial regression. This is consistent with previous findings that birth weight, gestational age, and weight-for-gestational age do not modify the effect of MMS on infant mortality (5). To reduce bias from confounding between the mediators and the outcome, the analysis controlled for maternal age (≤19, 20–24, 25–29, and ≥30 y), maternal baseline BMI (≤19, 20–24, 25–29, and ≥30 kg/m2), maternal baseline hemoglobin (≥11 g/dL and <11 g/dL), sex of the child, maternal marital status (married, cohabiting, and unmarried and not cohabiting), parity (0, 1, 2, and ≥3), wealth quartile, and maternal education (none, primary, secondary, or higher). All confounders were modeled as categorical variables to allow for nonlinear relations with the outcome, and missing indicators were used to account for missing confounder data. Mediation analyses were conducted as complete-case for each mediating variable.
The flexible weighting approach to mediation described by VanderWeele and Vansteelandt was used in all analyses; this method allowed the mediator–outcome relations to be modeled as curvilinear functions (11). SEs and CIs were calculated through use of 1000 bootstrapped simulations (12). For the analysis of the joint mediation by gestational age and weight-for-gestational age, we allowed for interactions between the two mediators on the outcome.
To further understand the pathways described by the mediation analyses, we performed regression analyses to determine whether relations existed between MMS and the mediators of interest, and between the mediators of interest and 6-wk infant mortality. We used ordinary least squares regression to determine the average effect of MMS on the continuous outcomes of birth weight, gestational age, and weight-for-gestational age. We used log-binomial regression to determine the relative risk of MMS on dichotomous outcomes of low birth weight (<2500 g), preterm birth (<37 wk), and SGA births (<10th percentile for weight-for-gestational age). The log-binomial regression controlled for the same set of confounders as the mediation analysis. In cases where the log-binomial models did not converge, modified Poisson models were used to provide consistent but not fully efficient estimates of the RR and CIs (13).
Sensitivity analyses were conducted to investigate the robustness of the results to the inclusion of twins and also under single imputation for missing confounder data. These sensitivity analyses used similar methods to those described earlier, but for the twins analysis we estimated SEs through use of clustered bootstrapping to account for correlations between siblings. Analyses were conducted with SAS software, Version 9.4 (SAS Institute).
Results
The parent trial enrolled 8428 pregnant mothers, of whom 8379 had birth outcome data. After excluding multiple births (n = 156), miscarriages (n = 102), stillbirths (n = 277), and losses to follow-up (n = 358), 7486 woman-child dyads remained in the study population (Figure 2). Table 1 presents baseline characteristics of the singleton live birth study cohort stratified by randomized MMS and IFA regimen. Maternal age, maternal baseline BMI, maternal baseline hemoglobin, sex of the child, marital status of the mother, parity, wealth quartile, and maternal education were balanced across study arms.
FIGURE 2.
Analytic population.
TABLE 1.
Characteristics of pregnant women randomly assigned to MMS or placebo included in the mediation analysis (n = 7486)1
| MMS (n = 3762) | Placebo (n = 3724) | |
|---|---|---|
| Covariate | n (%) | n (%) |
| Maternal age, y | ||
| <20 | 564 (15.0) | 608 (16.3) |
| 20–24 | 1507 (40.1) | 1452 (39.0) |
| 25–29 | 1028 (27.3) | 1008 (27.1) |
| ≥30 | 647 (17.2) | 633 (17.0) |
| Missing | 16 (0.4) | 23 (0.6) |
| Maternal BMI, kg/m2 | ||
| <22 | 869 (23.1) | 888 (23.8) |
| 22–24.9 | 1128 (30.0) | 1159 (31.1) |
| 25–29.9 | 996 (26.5) | 907 (24.4) |
| ≥30 | 297 (7.9) | 321 (8.6) |
| Anemia, Hb <11.0 g/dL | ||
| Yes | 2189 (58.2) | 2128 (57.1) |
| No | 1047 (27.8) | 1080 (29.0) |
| Missing | 526 (14.0) | 516 (13.9) |
| Marital status | ||
| Married | 2524 (67.1) | 2502 (67.2) |
| Cohabiting | 793 (21.1) | 764 (20.5) |
| Unmarried | 423 (11.2) | 423 (11.4) |
| Missing | 22 (0.6) | 35 (0.9) |
| Parity | ||
| 0 | 1673 (44.5) | 1633 (43.9) |
| 1 | 1056 (28.1) | 1038 (27.9) |
| 2 | 543 (14.4) | 577 (15.5) |
| ≥3 | 474 (12.6) | 451 (12.1) |
| Missing | 16 (0.4) | 25 (0.7) |
| Maternal education | ||
| None | 440 (11.7) | 398 (10.7) |
| Primary | 2461 (65.4) | 2521 (67.7) |
| Secondary | 647 (17.2) | 607 (16.3) |
| Higher | 200 (5.3) | 177 (4.8) |
| Missing | 14 (0.4) | 21 (0.6) |
MMS, maternal micronutrient supplementation.
Among all singleton live births, MMS provided a nonstatistically significant 10% reduction in risk of 6-wk infant mortality (RR: 0.90; 95% CI: 0.69, 1.19) (Supplemental Table 1). However, MMS reduced the risk of low birth weight births (RR: 0.78; 95% CI: 0.65, 0.93) and reduced the risk of SGA births (RR: 0.76; 95% CI: 0.66, 0.87). Although MMS did not reduce the risk of preterm birth (RR: 1.02; 95% CI: 0.92, 1.13), it did increase the average gestational age across the population by 1.2 d (95% CI: 0.2, 2.1). Further, we found that each 100 g increase in birth weight was associated with 15% (RR: 0.85; 95% CI: 0.84, 0.87) lower risk of mortality, and each week increase in gestation was associated with 10% reduction in the risk of mortality (RR: 0.90; 95% CI: 0.86, 0.93).
In mediation analyses of all infants, significant NIEs were determined for birth weight (P < 0.001), weight-for-gestational age alone (P < 0.001), and the joint effects of gestational age and weight-for-gestational age (P < 0.001). For the total population, birth weight accounted for 75% of the observed effect of MMS on 6-wk infant mortality (Table 2). The NIE, or the amount of the effect explained by the pathway through the mediator, showed a significant −0.4% (95% CI: −0.6%, −0.2%, P < 0.001) decrease in absolute risk of mortality as a result of MMS. The NDE, or the risk difference not explained by the pathway through the mediator, showed a −0.1% (95% CI: −0.1%, −0.1%) decrease in mortality as a result of MMS. The pathway through gestational age alone accounted for 35% of the total observed effect of MMS on 6-wk mortality and the pathway through weight-for-gestational age alone accounted for 22% of the effect (Table 2). When we considered the joint mediation of the effect of MMS through both gestational age and weight-for-gestational age and accounted for the interaction between the mediators, this fully explained the observed magnitude of the effect of MMS on 6-wk mortality. In fact, the NDE increased mortality. These results were consistent with the simple regressions of the effect of MMS on each mediator and the mediators on mortality. Among the singleton population, MMS had an effect on birth weight, gestational age, and weight-for-gestational age (Supplemental Tables 2and 3). There was also an observed effect of each mediator on 6-wk mortality controlling for confounders (Supplemental Table 4).
TABLE 2.
Mediation analysis results among all live births (n = 7486)1
| Mediator | NIE,2,3 % change in risk difference (95% CI) | NDE,3,4 % change in risk difference (95% CI) | Total effect,3 % change in risk difference (95% CI) | Proportion mediated, % |
|---|---|---|---|---|
| A. Birth weight (n = 7242) | −0.39 (−0.59, −0.19) | −0.13 (−0.14, −0.13) | −0.52 (−0.72, −0.32) | 74.6 |
| B. Gestational age (n = 7478) | −0.12 (−0.26, 0.01) | −0.22 (−0.23, −0.21) | −0.34 (−0.49, −0.21) | 35.3 |
| C. Weight-for-gestational age (n = 7221) | −0.10 (−0.17, −0.05) | −0.36 (−0.37, −0.35) | −0.47 (−0.53, −0.41) | 22.2 |
| D. Joint effects of gestational age and weight-for-gestational age (n = 7221) | −0.36 (−0.54, −0.18) | 0.00 (0.00, 0.00) | −0.35 (−0.53, −0.18) | Fully mediated |
NDE, natural direct effects; NIE, natural indirect effects.
NIEs were calculated with use of a causal mediation framework and estimated with a weighting-based counterfactual approach to mediation. The NIE is interpretable as the average risk difference of mortality comparing the marginal effect of the mediator when the population is exposed to multiple micronutrient supplements.
All analyses controlled for confounding by maternal age, maternal baseline BMI, maternal baseline hemoglobin, infant sex, maternal marital status, parity, maternal wealth quartile, and maternal education. The weights for the counterfactual averages were estimated through use of logistic regression. Counterfactual weighted averages were used to estimate risk difference estimates for the NIE, NDE, and total effects.
The NDE is interpretable as the average risk difference of mortality comparing the marginal effect of multiple micronutrient supplements when the mediator is fixed to its natural value under placebo. The proportion mediated indicates how much of the total effect passes through the mediator of interest.
Mediation of the effect of MMS stratified by the sex of the child
Among the singleton live births in the study, MMS decreased male 6-wk mortality by a nonstatistically significant 20% (RR: 0.80; 95% CI: 0.53, 1.19) and decreased female 6-wk mortality by a nonstatistically significant 10% (RR: 0.90; 95% CI: 0.55, 1.47).
We performed the same four sets of mediation analyses on the data stratified by sex. Among female infants, mediation by the joint effect of gestational age and weight-for-gestational age fully explained the decrease in mortality associated with MMS (Table 3). Significant NIEs were found for birth weight, gestational age alone (P < 0.001), weight-for-gestational age alone (P < 0.001), and the joint effect of gestational age and weight-for-gestational age (P < 0.001). Any observed direct effect of MMS on 6-wk mortality outside gestational age and weight-for-gestational age increased mortality. Looking at mediation through weight-for-gestational age and gestational age separately, gestational age alone mediated more of the effect than weight-for-gestational age.
TABLE 3.
Mediation analysis results among live births by sex1
| Mediator | NIE,2 % change in risk difference (95% CI) | NDE,2 % change in risk difference (95% CI) | Total effect,2 % change in risk difference (95% CI) | Proportion mediated, % |
|---|---|---|---|---|
| Female infants (n = 3570) | ||||
| A. Birth weight (n = 3493) | −0.45 (−0.67, −0.20) | 0.21 (0.20, 0.22) | −0.24 (−0.46, 0.00) | Fully mediated |
| B. Gestational age (n = 3566) | −0.21 (−0.32, −0.10) | 0.04 (0.04, 0.04) | −0.17 (−0.28, −0.07) | Fully mediated |
| C. Weight-for-gestational age (n = 3490) | −0.14 (−0.25, −0.03) | −0.05 (−0.05, −0.04) | −0.18 (−0.30, −0.07) | 74.6 |
| D. Joint effects of gestational age and weight-for-gestational age (n = 3490) | −0.38 (−0.60, −0.16) | 0.20 (0.19, 0.21) | −0.18 (−0.40, 0.03) | Fully mediated |
| Male infants (n = 3833) | ||||
| A. Birth weight (n = 3735) | −0.25 (−0.44, −0.06) | −0.42 (−0.43, −0.40) | −0.66 (−0.87, −0.46) | 37.3 |
| B. Gestational age (n = 3829) | −0.03 (−0.14, 0.09) | −0.53 (−0.55, −0.51) | −0.56 (−0.68, −0.43) | 5.0 |
| C. Weight-for-gestational age (n = 3731) | −0.06 (−0.10, −0.02) | −0.60 (−0.61, −0.59) | −0.66 (−0.71, −0.61) | 9.1 |
| D. Joint effects of gestational age and weight-for-gestational age (n = 3731) | −0.22 (−0.41, −0.02) | −0.44 (−0.46, −0.42) | −0.66 (−0.87, −0.45) | 33.9 |
NDE, natural direct effects; NIE, natural indirect effects.
All analyses controlled for confounding by maternal age, maternal baseline BMI, maternal baseline hemoglobin, maternal marital status, parity, maternal wealth quartile, and maternal education. The effects were estimated through use of logistic regression for binary outcomes.
For the male infants, the indirect effect through the joint mediators of gestational age and weight-for-gestational age was smaller than that for females (Table 3). However, there were significant NIEs for birth weight (P < 0.001), weight-for-gestational age (P < 0.001), and the joint effect of gestational age and weight-for-gestational age (P < 0.001). Only 34% of the total effect of MMS on mortality for male infants could be explained by the joint mediation of gestational age and weight-for-gestational age. For mediation by gestational age alone and by weight-for-gestational age alone, very little of the effect among males was explained by these mediators (5.0% and 9.1%, respectively).
We further investigated the sex-specific relations between MMS and the mediators (Supplemental Tables 2 and 3), and between the mediators and 6-wk mortality (Supplemental Table 4). Greater gestational age and weight-for gestational age reduced mortality by similar amounts for males and females. However, MMS had a differential effect on gestational age for females and males. For females, MMS increased gestational age by an average of 0.25 wk (95% CI: 0.06, 0.44) compared to 0.10 wk (95% CI: −0.09, 0.29) among males (Supplemental Figure 1).
Sensitivity analyses
We conducted a sensitivity analysis that included the 156 twins. The proportion mediated decreased; however, patterns remained consistent with the singleton analyses (Supplemental Table 5). In analyses of female infants including twins, the joint effect of gestational age and weight-for-gestational age fully mediated the effect of MMS on mortality, whereas among males only 18.4% of the effect was mediated through the joint effect (Supplemental Table 6). Use of single imputation to account for missing data did not significantly change the results of the singleton analyses (Supplemental Tables 7and 8).
Discussion
We found significant NIEs of MMS on mortality through birth weight and the joint effect of gestational age and weight-for-gestational age. In analyses among all infants, we found that a substantial proportion (75%) of any effect of MMS on 6-wk infant mortality was mediated by increased birth weight. Low birth weight is a consequence of both intrauterine growth restriction and preterm birth. In fact, further analyses demonstrated that increased gestational age alone explained over one-third of the observed effect of MMS, while improved weight-for-gestational age alone explained about 20% of the effect. However, a joint analysis of gestational age and weight-for-gestational age that allowed for potential interaction of these mediators showed full mediation of the effect of MMS on mortality among all infants. Further, we identified differences in the mediation pathways by infant sex. Among female infants, the effect of MMS on mortality was fully mediated by the joint effects of gestational age and weight-for-gestational age, whereas these pathways only accounted for 34% of the observed total effect of MMS in males.
Preterm birth and intrauterine growth restriction are leading causes of child mortality globally. Global mortality models suggest that complications from preterm birth account for the largest proportion of child deaths (18%) (14), and recent surveillance data from nearly 300,000 pregnancies globally show that 1 in 5 newborn deaths are the result of preterm birth complications (15). Preterm infants are estimated to have 6.8 (95% CI: 3.6, 13.1) times the risk of neonatal mortality, and SGA infants are nearly two times more likely to die in the first month of life compared to term, appropriate-for-gestational-age infants (16). Our work in Tanzania has similarly shown the excess mortality risk associated with being born too small or born too soon (17, 18).
There is evidence that MMS reduces the risk of preterm, low birth weight, and SGA births (4, 5). MMS may increase gestation length and support improved fetal growth by improving the immune function of the neonate. The hypothesis that MMS improves infant immune function is supported by a recent trial in The Gambia, which found that MMS in pregnancy enhanced infant antibody response to the first diphtheria-tetanus-pertussis vaccine (19). MMS also improves the maternal and fetal immune system by reducing inflammation from infection, resulting in improved metabolism, improved placental function, and reduced maternal hypertension (20–22). As a result, MMS is hypothesized to result in longer gestation and increased weight-for-gestational age that lead to subsequent improvements in mortality.
In line with this hypothesis, we identified significant NIEs of MMS on mortality through birth weight and the joint effects of gestational age and weight-for-gestational age. This suggests that postpartum mechanisms minimally, if at all, contribute to the relation of MMS with 6-wk mortality in the context of the Tanzania trial. Nevertheless, it has been proposed that MMS may increase survival by strengthening the micronutrient content of maternal breast milk (23, 24). Our analysis suggests that this mechanism does not play a major role in infant survival during the first 6 wk of life, although effects on post 6-wk mortality cannot be ruled out. Additional studies will be necessary to investigate whether there are longer-term survival gains of MMS mediated by improvements in breast milk micronutrient status.
There is evidence that MMS produces larger mortality benefits for female infants compared to male infants (5, 6). Our study supports the hypothesis that these sex-specific effects result from differences in mechanisms related to birth outcomes. When the Tanzania trial data were stratified by sex, gestational age and weight-for-gestational age fully explained the observed effect of MMS on 6-wk infant mortality among female infants. We also found a significant NIE of gestational age alone among females. However, among male infants, gestational age and weight-for-gestational age only mediated 33.9% of the effect of MMS on 6-wk mortality. There was no evidence of an NIE through gestational age alone among males. There are two potential reasons for the differences in the strength of these mediation pathways by sex. First, our data suggest that male infants were less responsive to MMS regarding gestational age and weight-for-gestational age. MMS increased gestational age on average by only 0.10 wk among male infants compared to 0.25 wk among female infants. In the Tanzania trial population, each week increase in gestational age was associated with a 15% reduction in the risk of mortality and the strength of this association did not differ by sex. Second, the causes and timing of death during infancy may be different between males and females. However, no published data directly describe sex-specific causes of death in the neonatal period for infants in LMICs. We hypothesize that complications from preterm birth and SGA may be more common causes of neonatal mortality for female infants. In contrast, male infants may be at higher risk of death as a result of infection after birth. Observational data suggest that neonatal infection may be more prevalent among male infants compared to female infants (25). A third possible mechanism driving the sex differential is that male infants may be more affected by birth asphyxia because of their larger birth size. However, we did not find specific evidence in support of this hypothesis as there was a similar magnitude of association of weight-for-gestational age and gestational age with mortality for male and female infants (Supplemental Table 4).
It is important to note that in the Tanzania trial we did not observe a statistically significant total effect of MMS on mortality overall, nor among female or male infants. Nevertheless, we did identify statistically significant NIEs overall and for both sexes. Mediation analyses can be used to investigate pathways when there are small or nonsignificant total effects, particularly in cases with well-established a priori hypotheses (26). Our hypothesis that the effect of MMS on mortality is mediated by increased gestation duration and improved intrauterine growth is well supported by the literature (4, 5, 16, 17). Further, evidence from a recent meta-analysis of MMS on neonatal mortality consistently shows greater beneficial effects of MMS on mortality for female infants compared to male infants (5). This is in contrast to the Tanzania trial where the total effect in males was stronger than females, although neither were statistically significant. As a result, we may be observing normal statistical variation in the Tanzania data (i.e., we observed the effect in males by chance). Alternatively, there may be different effects of MMS on male infant mortality in the context of Tanzania compared to other MMS trials in South Asia and sub-Saharan Africa (4, 5); however, this seems unlikely.
There are also a few limitations to this study. The parent trial used a multiple micronutrient formulation different from the standard UNIMMAP formulation; the supplement did not include minerals other than iron. Another limitation of this study was the use of LMP to estimate gestational age. LMP is vulnerable to inaccurate recall and digit preference, and because of the assumption that ovulation occurs at day 14, it often overestimates gestational age. Nevertheless, the nondifferential measurement error in LMP would underestimate the amount of mediation we would expect to observe. Another limitation inherent to mediation analyses is the potential for unmeasured confounding of the mediator-outcome relation. However, we adjusted for major confounders such as maternal age, maternal baseline BMI, maternal baseline hemoglobin, sex of the child, maternal marital status, parity, wealth quartile, and maternal education. Residual confounding is also possible between the mediator and outcome within categories of the measured confounders included in the analysis.
This study has several implications. Foremost, it gives insight to the potential mechanisms leading to sex-specific effects of MMS. In females, the effect of MMS on mortality appears to operate entirely through increased gestation duration and intrauterine growth. As a result, these outcomes as well as birth weight appears to be an appropriate surrogate indicator for an effect of MMS on mortality among female infants in studies and programs. In contrast, if there is an effect of MMS on 6-wk infant mortality among males, MMS appears to work through mechanisms beyond duration of gestational age and increased weight-for-gestational age. It is important to note that this analysis does not call for differential provision of MMS by infant sex. This study provides additional support for implementing MMS as a public health program in LMICs to contribute to reaching the Sustainable Development Goal for child mortality by 2030.
Supplementary Material
ACKNOWLEDGEMENTS
The authors’ responsibilities were as follows–MKQ, ERS, PLW, WU, GM, WWF, and CRS: designed research; WU and GM: conducted research; MKQ and JS: analyzed data; MKQ and JS: wrote the paper; MKQ: primary responsibility for final content; and all authors: read and approved the final manuscript.
Notes
Supported in part by the National Institutes of Health (NIH) predoctoral training grant T32 AI007358, and was a secondary analysis of data from a study supported by a grant from the National Institute of Child Health and Human Development (NICHD R01 37701).
Author disclosures: MKQ, ERS, PLW, WU, JS, GM, WWF, and CRS, no conflicts of interest.
Supplemental Tables 1–8 and Supplemental Figure 1 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/.
Abbreviations used: IFA, iron-folic acid; LMIC, low- and middle-income country; LMP, last menstrual period; MMS, maternal micronutrient supplementation; NDE, natural direct effect; NIE, natural indirect effect; SGA, small-for-gestational-age.
References
- 1. WHO, Food and Agriculture Organization of the UN. Vitamin and mineral requirements in human nutrition. 2nd ed, Bangkok, Thailand; 2004. [Google Scholar]
- 2. Gernand AD, Schulze KJ, Stewart CP, West KP Jr., Christian P. Micronutrient deficiencies in pregnancy worldwide: health effects and prevention. Nat Rev Endocrinol. 2016;12:274–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Black RE. Micronutrients in pregnancy. Br J Nutr. 2001;85:S193–7. [DOI] [PubMed] [Google Scholar]
- 4. Keats EC, Haider BA, Tam E, Bhutta ZA. Multiple-micronutrient supplementation for women during pregnancy. Cochrane Database Syst Rev. 2019;3:CD004905. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Smith ER, Shankar AH, Wu LSF, Aboud S, Adu-Afarwuah S, Ali H, Agustina R, Arifeen S, Ashorn P, Bhutta ZA et al.. Modifiers of the effect of maternal multiple micronutrient supplementation on stillbirth, birth outcomes, and infant mortality: a meta-analysis of individual patient data from 17 randomised trials in low-income and middle-income countries. Lancet Glob Health. 2017;5:e1090–e100. [DOI] [PubMed] [Google Scholar]
- 6. West KP Jr., Shamim AA, Mehra S, Labrique AB, Ali H, Shaikh S, Klemm RD, Wu LS, Mitra M, Haque R et al.. Effect of maternal multiple micronutrient vs iron-folic acid supplementation on infant mortality and adverse birth outcomes in rural Bangladesh: the JiVitA-3 randomized trial. JAMA. 2014;312:2649–58. [DOI] [PubMed] [Google Scholar]
- 7. Lee YC, Chan SH, Ren EC. Asian population frequencies and haplotype distribution of killer cell immunoglobulin-like receptor (KIR) genes among Chinese, Malay, and Indian in Singapore. Immunogenetics. 2008;60:645–54. [DOI] [PubMed] [Google Scholar]
- 8. Fawzi WW, Msamanga GI, Urassa W, Hertzmark E, Petraro P, Willett WC, Spiegelman D. Vitamins and perinatal outcomes among HIV-negative women in Tanzania. N Engl J Med. 2007;356:1423–31. [DOI] [PubMed] [Google Scholar]
- 9. VanderWeele TJ. Explanation in causal inference: methods for mediation and interaction. Oxford University Press; 2015. [Google Scholar]
- 10. Stirnemann J, Villar J, Salomon LJ, Ohuma E, Ruyan P, Altman DG, Nosten F, Craik R, Munim S et al.. International estimated fetal weight standards of the INTERGROWTH-21(st) Project. Ultrasound Obstet Gynecol. 2017;49:478–86. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. VanderWeele TJ, Vansteelandt S.. Mediation analysis with multiple mediators. Epidemiol Methods. 2014;2:95–115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Imai K, Keele L, Tingley D. A general approach to causal mediation analysis. Psychol Methods. 2010;15:309–34. [DOI] [PubMed] [Google Scholar]
- 13. Zou G. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol. 2004;159:702–6. [DOI] [PubMed] [Google Scholar]
- 14. Liu L, Oza S, Hogan D, Chu Y, Perin J, Zhu J, Lawn JE, Cousens S, Mathers C, Black RE. Global, regional, and national causes of under-5 mortality in 2000–15: an updated systematic analysis with implications for the Sustainable Development Goals. Lancet. 2016;388:3027–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Ahmed I, Ali SM, Amenga-Etego S, Ariff S, Bahl R, Baqui AH, Begum N, Bhandari N, Bhatia K, Bhutta ZA et al.. Population-based rates, timing, and causes of maternal deaths, stillbirths, and neonatal deaths in south Asia and sub-Saharan Africa: a multi-country prospective cohort study. Lancet Glob Health. 2018;6:e1297–e308. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Katz J, Lee ACC, Kozuki N, Lawn JE, Cousens S, Blencowe H, Ezzati M, Bhutta ZA, Marchant T, Willey BA et al.. Mortality risk in preterm and small-for-gestational-age infants in low-income and middle-income countries: a pooled country analysis. Lancet. 2013;382:417–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Sania A, Smith ER, Manji K, Duggan C, Masanja H, Kisenge R, Msamanga G, Urassa W, Fawzi W. Neonatal and infant mortality risk associated with preterm and small for gestational age births in Tanzania: individual level pooled analysis using the intergrowth standard. J Pediatr. 2018;192:66–72 e4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Muhihi A, Sudfeld CR, Smith ER, Noor RA, Mshamu S, Briegleb C, Bakari M, Masanja H, Fawzi W, Chan GJ. Risk factors for small-for-gestational-age and preterm births among 19,269 Tanzanian newborns. BMC Pregnancy Childbirth. 2016;16:110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Okala SG, Darboe MK, Sosseh F, Sonko B, Faye-Joof T, Prentice AM, Moore SE. Impact of nutritional supplementation during pregnancy on antibody responses to diphtheria-tetanus-pertussis vaccination in infants: a randomised trial in The Gambia. PLoS Med. 2019;16:e1002854. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Romero R, Chaiworapongsa T, Espinoza J. Micronutrients and intrauterine infection, preterm birth and the fetal inflammatory response syndrome. J Nutr. 2003;133:1668S–73S. [DOI] [PubMed] [Google Scholar]
- 21. Christian P. Micronutrients, birth weight, and survival. Annu Rev Nutr. 2010;30:83–104. [DOI] [PubMed] [Google Scholar]
- 22. Priliani L, Prado EL, Restuadi R, Waturangi DE, Shankar AH, Malik SG. Maternal multiple micronutrient supplementation stabilizes mitochondrial DNA copy number in pregnant women in Lombok, Indonesia. J Nutr. 2019;149:1309–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Allen LH. Multiple micronutrients in pregnancy and lactation: an overview. Am J Clin Nutr. 2005;81:1206S–12S. [DOI] [PubMed] [Google Scholar]
- 24. Supplementation with Multiple Micronutrients Intervention Trial Study Group, Shankar AH, Jahari AB, Sebayang SK, Aditiawarman, Apriatni M, Harefa B, Muadz H, Soesbandoro SD, Tjiong R et al.. Effect of maternal multiple micronutrient supplementation on fetal loss and infant death in Indonesia: a double-blind cluster-randomised trial. Lancet. 2008;371:215–27. [DOI] [PubMed] [Google Scholar]
- 25. Seale AC, Blencowe H, Zaidi A, Ganatra H, Syed S, Engmann C, Newton CR, Vergnano S, Stoll BJ, Cousens SN et al.. Neonatal severe bacterial infection impairment estimates in South Asia, sub-Saharan Africa, and Latin America for 2010. Pediatr Res. 2013;74(Suppl 1):73–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Agler R, De Boeck P. On the interpretation and use of mediation: multiple perspectives on mediation analysis. Front Psychol. 2017;8:1984. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.