Maternal vitamin B12 in pregnancy and risk of preterm birth and low birth weight: A systematic review and individual participant data meta-analysis

Study inclusion criteria

We included studies that assessed the association between maternal B12 in serum or plasma during pregnancy and birth weight or gestational age at delivery.

Only studies of longitudinal cohort design were eligible for this review. To be eligible, information on birth weight had to be registered at birth and could not be retrospectively reported and length of gestation, in completed days or weeks, had to be estimated by either ultrasound or last menstrual period, or a combination of the two. Studies where B12 was measured after conception and prior to delivery were eligible. If a study was designed to evaluate women or offspring with a specific condition (e.g. preeclampsia or congenital malformations), and there was a marked overrepresentation of participants with such a condition, that study was excluded. Studies with fewer than 50 participants were not considered. Given the need to collaborate with authors of the original studies, we included only those studies published in 1998 or later.

Search methods

The electronic literature search was constructed by the first author (TR) and a librarian trained in medical database searches, and conducted in PubMed, Scopus, Web of Knowledge, EBSCO-host (CINAHL) and OvidSP (MEDLINE, EMBASE and GLOBAL HEALTH); last accessed August 2015. No language restriction was applied. The reference lists of all studies read in full-text were hand searched to find additional eligible studies. Web Appendix 1 provides complete information on the electronic searches.

Data collection

Electronic literature searches were carried out by the first author (TR). Duplicates were removed and eligibility of all references evaluated by screening of the titles and abstracts by the first author (TR). All potentially eligible studies were read in full-text and assessed for inclusion independently by two authors (TR and KRR). A handsearch of reference lists was done independently by two authors (TR and KRR or MJT). When multiple reports from the same study were found, we used the most complete report.

Risk of bias was independently assessed by two authors (TR and MJT) based on a modified version of the Newcastle-Ottawa Scale (range 0-7) (14). Disagreements were resolved by consulting a third reviewer (KRR). We defined high risk of bias as a score of four or less, and moderate to low risk was defined as scores five through seven.

Authors from all eligible studies were contacted to obtain IPD, each research group being approached at least three times. IPD was received without personal identification. For studies where IPD could not be shared, authors were asked to provide results from pre-specified reanalyzes of their data. When neither IPD nor reanalyzes could be retrieved, relevant estimates were extracted from the publications.

Variables

The main exposure of interest was vitamin B12 levels in maternal serum or plasma. We calculated trimester-specific standard deviation (SD) scores based on studies providing IPD and reanalyzed aggregate data. Analyses were performed for B12-deficiency pre-defined as <148 pmol/L (15), and B12 tertiles constructed on the basis of included individual data; <148 pmol/L (tertile 1), 148-216 pmol/L (tertile 2), and >216 pmol/L (tertile 3).

The three pre-defined main outcomes were: birth weight as a continuous measure in grams, LBW (birth weight <2,500 g) and small-for-gestational-age (SGA; birth weight SD score <10^th centile) (1). Birth weight SD score was calculated using gestational age at delivery and sex-specific reference standards published by the INTERGROWTH 21^st Project (16). We assumed birth weight SD score to serve as a proxy of fetal growth, and defined SGA as a proxy of restricted fetal growth. Outcomes related to length of gestation were gestational age at delivery (days) and preterm birth (gestational age at delivery <37 weeks).

Three main confounders were identified based on a priori assumptions of confounding factors, availability of data and exploration of effect of covariates on outcome and exposure: maternal age (continuous), pre-pregnancy or pregnancy body mass index (BMI, continuous) and parity (nulliparous versus primiparous or multiparous). Maternal weight was used when information on BMI was unavailable. Also, we considered smoking habits (smoking versus not smoking during pregnancy) and highest completed education (completed high school, equal to 13 years of education, versus not completed high school).

Details of eligible studies

Studies included in the meta-analyses are described in Table 1; details of the eligible studies not included are presented in Web Table 1 (36–39). Of the included studies, one was conducted in North America (34), nine in Europe (18,19,22,25–28,31,32), one in Africa (30), one in Oceania (24), and six in Asia (20,21,23,29,33,35). The number of pregnancies studied ranged from 84 to 5,641. B12 was measured during the first trimester in seven studies (19,22,23,28,31–33), during the second trimester in 15 studies (18–24,26–29,31–33,35), and during the third trimester in 12 studies (18,20,21,23,25,27,29,30,32–35). Mean (SD) B12 concentrations in the first, second and third trimester were 219.8 (128.2), 187.8 (91.3) and 188.7 (82.5) pmol/L, respectively. Preterm deliveries were excluded from four studies (25,26,31,33).

Key maternal and newborn characteristics of the included studies are presented in Table 2. B12-deficiency was identified in 0% to 69% of pregnancies (median 33%). The incidence of LBW ranged from 0% to 33% (median 6%), preterm births from 4% to 14% (median 8%), and SGA from 5% to 32% (median 11%). Higher maternal weight was associated with lower maternal B12; one SD higher maternal BMI or weight was associated with an 11 pmol/L decrease in B12 (95% CI -15, -7).

Systematic review

Meta-analysis of maternal B12 in relation to birth weight and LBW

In the meta-analysis, we found no evidence of a linear association between B12 and birth weight (Figure 2): The adjusted estimate was 5.1 g increase in birth weight per SD increase in B12 (95% CI -10.9, 21.0; I²=30%).

Open in a separate window

Figure 2

Forest plot presenting the association between B12 and birth weight

Meta-analysis of studies of the association between vitamin B12 and birth weight after adjustment for maternal age, parity and body mass index or weight. Effect estimates are expressed as change in birth weight per one standard deviation increase of vitamin B12 (i.e. mean difference). CI, confidence interval; No, number pregnancies.

Subgroup and sensitivity analyses are presented in Web Table 3. Stratification by country income showed that there was an association between B12 and birth weight in low- and middle-income countries, but not in high-income countries. Heterogeneity among the studies was explained largely by country income level and maternal BMI or weight. Excluding a study that used late-pregnancy BMI (29), instead of pre-pregnancy or early pregnancy BMI/weight in the other studies, reduced the heterogeneity from I²=30% to I²=13% (results not presented). One study reported an association between B12 and birth weight that greatly deviated from the other studies (33). Excluding this study did not notably change the effect estimate, but resulted in a modest reduction in heterogeneity (from I²=30% to I²=21%; results not presented). Sensitivity analyses excluding each of the included studies one by one, and excluding studies only evaluating newborns born at term, did not meaningfully alter the association between B12 and birth weight (results not presented).

Results for categories of B12 supported our main results. Neither B12-deficiency nor B12 tertiles were associated with birth weight (see Web Table 4).

B12-deficiency was associated with a 15% (95% CI 1%, 31%; I²=5%) increased risk of LBW (Figure 3A).

Open in a separate window

Figure 3A

Forest plot presenting the association between B12-deficiency and the risk of low birth weight

Meta-analysis of studies of the association between vitamin B12-deficiency and the risk of low birth weight after adjustment for maternal age, parity and body mass index or weight. Effect estimate expressed as risk ratio of the outcome comparing B12 deficient to non-deficient. CI, confidence interval; No, number pregnancies.

The funnel plot of B12 and birth weight indicated low risk of publication bias (see Web Figure 1).

Since birth weight may be regarded as a summary measure of fetal growth and gestational age, we further performed analyses to assess a possible influence of B12 on these factors.

Meta-analysis of maternal B12 in relation to length of gestation and preterm birth

The analyses did not support a linear association between maternal B12 levels with length of gestation in days (0.1 days (95% CI -0.2, 0.3; I²=0%) per SD increase of B12). However, increasing levels of B12 were associated with a reduced risk of preterm birth (RR 0.89 (95% CI 0.82, 0.97; I²=0%) per SD increase in B12; Web Figure 2). Accordingly, B12-deficiency in pregnancy was associated with a 21% increased risk of preterm birth (95% CI -1%, 49%; I²=20%); Figure 3B).

Open in a separate window

Figure 3B

Forest plot presenting the association between B12-deficiency and the risk of preterm birth

Meta-analysis of studies of the association between vitamin B12-deficiency and the risk of preterm birth after adjustment for maternal age, parity and body mass index or weight. Effect estimate expressed as risk ratio of the outcome comparing B12 deficient to non-deficient. CI, confidence interval; No, number pregnancies.

The association between B12 and preterm birth was similar within all subgroup and sensitivity analyses, although there was a loss of precision in these subgroup analyses due to smaller sample sizes (see Web Table 5).

Strengths and limitations

A strength of this study is the use of IPD and reanalyzed data. Due to substantial heterogeneity in the published analyses, a traditional meta-analysis could not answer our research questions. Incomplete or selective reporting may reduce the replicability of studies and distort the literature (40). This is illustrated by comparing the findings of this review with those of a recently published systematic review by Sukumar et al. with traditional meta-analysis on the association between B12 and birth weight (9). That study reported an odds ratio of 1.70 (95% CI 1.16, 2.50; I²=84%) of the association between “low B12” and “adverse birth weight”. A more moderate association was found in the present study in a comparable analysis of B12-deficiency in relation to LBW (RR 1.15 (95% CI 1.01, 1.31); I²=5%). One reason for the discrepant results may be that Sukumar et al. depended solely on data presented in the published reports and were unable to include results reported as being “insignificant”; for instance, from the largest individual study in the present review (19). The comparable meta-analysis in the present review included roughly ten times as many pregnancies as the meta-analysis in Sukumar’s review. Additionally, of eight individual results included in Sukumar’s meta-analysis, five evaluated most of the same women from a single original study, exaggerating the influence of a single, outlying study (8,23). By collecting IPD and requesting reanalyzes from contributing studies, we were able to standardize the analyses across most of the included studies, thereby reducing heterogeneity and facilitating interpretation of results. Compared with the review by Sukumar et al. that presented meta-analyses with high levels of heterogeneity (I²-scores from 74% to 98% in the primary analyses), the present study had I²-scores between 0% and 30% in the primary analyses. Additionally, the present study enabled conduction of subgroup and sensitivity analyses, along with more complete adjustment for important confounders (e.g. maternal weight).

We included 94% of all eligible participants, permitting an unbiased summary of the published literature. Given the relative large number of included subjects, we had increased power to evaluate findings reported with low precision in individual studies. We tested the stability of our findings with a broad range of sensitivity analyses.

Another strength was that our analyses were not post-hoc, but followed a detailed protocol. We performed a thorough literature search without language restrictions, and systematically reviewed all eligible studies.

There are several limitations. Unpublished studies were not considered for this review, which could potentially skew estimates. However, a funnel plot did not suggest publication bias. We were unable to include four eligible studies (777 observations, 6% of all observations). Given the small number of observations, it is unlikely that inclusion of these remaining studies would have importantly influenced our main results.

Our approximations of fetal growth and restricted fetal growth by use of gestational age and sex specific birth weight charts is suboptimal, as these outcomes are ideally estimated using serial ultrasound measurements during pregnancy (41). Furthermore, we did not have sufficient data at hand to evaluate the possible implications of low levels of B12 during different periods in pregnancy in the same woman. Sensitivity analyses stratified by trimester of B12 measurement across studies, however, did not reveal important variation in the association between B12 and the outcomes of interest.

Importantly, B12-deficiency may be a proxy for inadequate nutritional status, and it is possible that some of our findings are related to nutritional status, not specifically to B12. A predominantly plant-based diet is low in B12, but also other nutrients, such as vitamin D and zinc, which to some degree may be associated with preterm birth (42–44). We did not have information on dietary intake or blood levels of these nutrients. Nutritional status could explain the present finding of an association between B12 and birth weight in low- and middle-income countries but not high-income countries. However, lower vitamin B12 levels were associated with higher risk of preterm birth irrespective of country income. It seems less likely that nutritional status can fully explain this finding.

Mixing of effects is inherent in observational studies, and residual confounding cannot be ruled out. We emphasize that our study report associations, and that causal effects must be explored through trials (see below). Reassuringly, we found little discrepancy in the pooled results of adjusted main models as compared to extended adjusted models (i.e. additionally adjusting for maternal education and smoking habits).

Possible explanation of findings

Low birth weight is a result of preterm birth, of being born small at term, or a combination of the two (45). While we found an increased risk of preterm birth and LBW among B12-deficient women, there was little evidence that maternal B12 levels influenced offspring birth weight SD score or SGA status. It seems more likely that the observed higher risk for LBW in B12-deficient women can be explained by preterm birth rather than by reduced fetal growth.

Higher B12 was associated with higher birth weight in low- and middle-income countries, but not in high-income countries. Four of the five studies included in the low- and middle-income group were performed in an Indian population. Therefore, generalization of these results to low- or middle-income countries outside India should be treated with caution. Indian women generally have lower dietary intake of B12, due to a mainly vegetarian diet, making them susceptible to B12-deficiency (46). Additionally, Indian newborns are among the smallest in the world (45). Our findings suggest that pregnancies already at greatest risk of giving birth to small newborns were the ones most vulnerable to low levels of B12.

The association between B12 and the risk of preterm birth was consistent across studies in both high-income and low- and middle-income countries, and generalization to countries not studied may be feasible.

In line with our findings, maternal obesity has been associated with B12-deficiency in several populations (47,48). It is hypothesized that this association is due to altered fat distribution and metabolism in the overweight compared with normal weight (47). Maternal weight is positively correlated with newborn weight (49), and failure to adjust for maternal weight may underestimate a positive association between B12 and birth weight.

Potential mechanism of action

Preterm birth may be categorized into spontaneous and medically indicated, with varying etiologies (50). Unfortunately, information on spontaneous versus medically indicated preterm births were not available to us. Medically indicated preterm birth are most commonly caused by severe preeclampsia or severely restricted fetal growth (51). Our findings do not support maternal level of B12 to be associated with fetal growth. Maternal B12 may, however, be associated with risk of preeclampsia, potentially through homocysteine, but reports are discrepant (52–54). The rate of medically indicated preterm births is higher in high-income countries than in low- and middle-income countries (55). In analysis stratified by country income, we found similar associations between B12 and risk of preterm birth in low-, middle- and high-income countries. Still, this finding does not link B12 to specific etiologies of preterm birth, which is a topic that deserves further studies.

It is possible that supplementation of B12 or folic acid, with a subsequent reduction of homocysteine, increases birth weight and length of gestation. However, a Cochrane review concluded that supplementation of folic acid during pregnancy did not reduce risk of either preterm birth or LBW (60). Two small (256 pregnancies and 68 pregnancies) randomized controlled trials of B12 supplementation during pregnancy reported on birth weight and length of gestation (61,62). Both observed higher B12 plasma levels in the supplemented group compared with the control group, but no reduction in homocysteine levels. No differences were observed in birth weight, length of gestation, or frequency of LBW births or preterm births in the supplemented group compared with the control group in either study (C Duggan, Harvard University, personal communication, 2015) (61,62). However, the studies were not powered to detect small but meaningful differences in preterm birth.

Context

There are 15 million preterm births and 20 million low birth weight births globally each year (1). The greatest burden of LBW is found in South Asia, while preterm birth is highest in Africa (1). Preterm birth is the leading cause of neonatal deaths (1). In the era of The Millennium Development Goals (1990-2015), post-neonatal under-five mortality rate was reduced by 58% (2). Reduction in neonatal mortality was less pronounced (47%) (2). Prevention of preterm birth is thus a key strategy to reduce neonatal deaths and reach the new target of under-five year mortality of 25 per 1,000 live births by 2030, down from 43 per 1,000 in 2015 (2).

Our systematic review was not designed to study the prevalence of B12-deficiency during pregnancy. However, this condition was common in the studies in our review, and comparable to a systematic review of B12-deficiency during pregnancy (9). A large group of women are thus affected by a potential preventable risk of preterm birth.

We would like to acknowledge the following persons for their help in order to provide the necessary data for the present study: Professor Caroline H D Fall, MRC Lifecourse Epidemiology Unit, University of Southampton, UK; Dr Simon Wheeler, nutritionist at the MRC Epidemiology Unit, University of Cambridge; Professor Anura V Kurpad, Professor of Physiology, St. John’s Medical College, India; and Dr Nithya Sukumar, research fellow at Warwick Medical School, University of Warwick, UK. Also thanks to Professor Christopher Duggan, Professor in the Department of Nutrition, Harvard University, USA, for providing necessary unpublished results from a previous trial. Lastly, we would like to acknowledge Janis Glover, Academic Librarian at the Medical Library, Yale University, USA, who helped develop the search strategy.

Funding source: This work was supported by the Norwegian University of Science and Technology as part of the Ph.D. in Medicine-programme. The funding source was not involved in the design of the study, statistical analysis and results interpretation. The researchers were independent from the funders.