Thirteen trials were included in the review, eight RCTs, one factorial RCT and four cluster RCTs. All the trials were classified as having a low or no risk of bias. Trials varied in whether they adjusted for confounders in their analysis and which confounders were used. The number of participants in the included trials ranged from 100 to 31,290 women.
Low birth weight: The risk of having a low birth weight infant was statistically significantly lower in the multi-micronutrient supplementation group compared to placebo (RR 0.81, 95% CI 0.73 to 0.91; four trials, n=6,097 women) and iron-folic acid supplementation (RR 0.83, 95% CI 0.74 to 0.93; 10 trials, n=29,889 women).
Other outcomes: There was a statistically significant increase in the birth weight of infants born to mothers in the multi-micronutrient supplementation group compared to the iron-folic acid supplementation group (mean difference 54g, 95% CI 36 to 72; eight trials, n=28,033 women) but not compared to placebo (mean difference 31g, 95% CI: -5 to 68; five trials, n=5,727 women). There was no statistically significant difference between the multi-micronutrient group and either comparator for risk of pre-term birth, having a small-for-gestational-age infant or infant gestational age.
There was moderate heterogeneity for the low birth weight analyses (I2 55% to 64%) but not for pre-term birth (I2 was not reported for the other outcomes). In a subgroup analysis of timing of initiation of supplementation before 20 weeks gestation, the difference between multi-micronutrient and iron-folic acid supplementation was no longer statistically significant, but in a subgroup of trials where initiation was after 12 weeks gestation, the results were similar to the main analysis. In a subgroup of trials of women with known HIV infection, there was no significant difference between the multi-micronutrient and iron-folic acid groups in the risk of having a low birth-weight infant.
The authors stated that visual inspection of the funnel plot suggested publication bias: trials of small effect size showing both positive and negative results may have been missed.