Prenatal folic acid and risk of asthma in children: a systematic review and meta-analysis

Search strategy

The study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (18). See “supplemental data – protocol” under “Supplemental data” in the online issue for our protocol. With the assistance of a research librarian, we searched PubMed (http://www.ncbi.nlm.nih.gov/pubmed), Medline (http://www.ncbi.nlm.nih.gov/), Embase (http://www.embase.com), Cumulative Index to Nursing and Allied Health Literature (http://www.ebscohost.com/academic/cinahl-plus-with-full-text), Cochrane (http://www.thecochranelibrary.com/view/0/index.html), Web of Science (http://thomsonreuters.com/web-of-science/), Population Information Online (http://www.popline.org/), and Education Resources Information Center (http://eric.ed.gov/) for the period from the inception of the database through February 2013 by using the term folic acid and its variants (folate, folacin, and vitamin B-9) combined with terms specific to allergic and respiratory outcomes (eg, asthma, respiratory, wheeze, reactive airway, atopy, and allergy) and their associated Medical Subject Headings. To ensure maximum sensitivity, no language or other limits were set. See “supplemental data – search strategies” under “Supplemental data” in the online issue for full search strategies for each database. We scanned reference lists of all included studies as well as those of relevant systematic review articles returned in the search, and articles with potentially relevant titles were pulled for screening against inclusion criteria.

Inclusion criteria

To be included in the systematic review, studies had to meet the following criteria: 1): be a randomized controlled trial, cohort, case-control, or cross-sectional study; 2) report the exposure of natural food folate intake, folic acid intake from fortified foods, total folate intake from foods (eg, dietary folate equivalents), folic acid intake from supplements, or maternal or cord blood serum, plasma, or red blood cell folate concentrations; 3) have an exposure timing during the periconceptional period or during pregnancy; 4) provide results on at least one allergic or respiratory outcome; and 5) include an evaluation of the direct association between folic acid exposure and one of the outcomes of interest.

Two independent reviewers (KSC and AMC) reviewed titles and, when available, the abstracts of search results and identified potentially relevant articles (Figure 1). The same authors independently reviewed the full text of each potentially relevant article against inclusion criteria. There were no discrepancies between authors on articles meriting inclusion.

Data extraction

One reviewer (AMC) extracted data into a piloted form, and the extracted data were independently cross-checked for accuracy with the original study by 2 additional reviewers (KSC and JM). The following elements were extracted: citation, study design (study type, study time period, study location, and study name), population (the number of study participants), exposure (type, measure, range, cutoffs, and methods of assessment), outcomes [type (eg, asthma, wheezing, allergy, eczema, and atopic dermatitis), how defined or measured, and timing], and results [the adjusted RR (aRR)⁵, adjusted OR (aOR), or adjusted prevalence ratio that reflected the greatest degree of control for confounders, CIs, and covariates adjusted for in the analysis].

Risk of bias and meta-analysis

We performed meta-analyses on articles with similar exposures and outcomes with Comprehensive Meta-Analysis Software (version 2; Biostat). Only the subset of studies of periconceptional folic acid exposure (month before and through the first trimester) and risk of asthma or wheezing was large enough to perform a meta-analysis of risk (criteria defined as ≥3 studies of sufficient quality). We measured the summary effect size for asthma alone and performed a sensitivity analysis with asthma combined with wheezing, which is sometimes used as a proxy for asthma. We combined the data from cohort studies and nested-case control studies by using ORs to approximate RRs. We used a random-effects model in our analyses. The risk estimate with the greatest control for confounders from each study was used in all analyses; see Supplemental Table 1 under “Supplemental data” in the online issue for a list of confounders adjusted for in each study. When risk estimates were provided for multiple exposure time points within the periconceptional period, exposure timing was prioritized in the following order for inclusion in the quantitative analysis (or meta-analysis), to prioritize exposure time periods closest to the DNA methylation programing in the early embryo: periconceptional, prepregnancy, and first trimester.

Statistical heterogeneity was tested with the Q statistic and evaluated by using the I² statistic (19), which provides an estimate of the proportion of the variance that is not attributable to random error. Heterogeneity was considered to be substantial if I² was greater than the conventional value of 50%. Publication bias was assessed by a visual inspection of funnel plots. The trim-and-fill analysis of Duval and Tweedie (20) was used to investigate the potential influence that unpublished studies could have on the summary estimates. P < 0.05 was considered statistically significant in all analyses.

Two authors (KSC and YPQ) independently graded each study included in the meta-analysis by using the Newcastle Ottawa Scale (NOS) as recommended by the Cochrane Non-Randomized Studies Methods Working Group (21). Studies assessed by using the NOS can receive a maximum score of 9. For our analyses, studies with a score ≥6 were considered to be of sufficient quality to include in the calculation of pooled risk estimates.

Outcomes: asthma or wheezing in childhood

Outcomes: other allergic (eg, atopy, eczema, and atopic dermatitis)

Outcomes: asthma or wheezing in childhood

Outcomes: other allergic (eg, atopy, eczema, and atopic dermatitis)

Any time or throughout pregnancy

Studies that examined folate at any time in pregnancy (eg, not stratified by the trimester of use) or throughout pregnancy are detailed in Tables 2 and 3 (see Supplemental Table 1 under “Supplemental data” in the online issue for additional details). Three of the studies showed no association (25, 33, 34) between folic acid and any of the outcomes assessed, whereas one study (24) showed slight increased risk between folic acid supplement use (compared with no use) and wheezing at 6–18 mo of age (aRR: 1.07; 95% CI: 1.02, 1.12) (24).

Limitations of current literature

One of the significant limitations of the current literature is the lack of randomized clinical trials because of ethical issues that would be involved in conducting such a study. Randomization of exposure is critical for the assessment of the effect of exposure to multivitamin supplements and folic acid supplements because the women who use folic acid supplements have different health behaviors and characteristics than do women who do not take supplements. Håberg et al (23) observed that supplement users were more likely to have atopic conditions, more education, lower smoking rates, and other health-behavior differences. Wald and Morris (36) published a reply to the study of Håberg et al (23) and argued that the weak association shown in the study was unlikely to be causal and more likely a result of confounding because women who are more health conscious are both more likely to take folic acid supplements and to report respiratory problems in their children. Other studies also showed that maternal folic acid supplement use was associated with maternal self-report of asthma or allergy (22, 26, 35). Several studies reported that children with asthma, wheezing, or atopy were significantly more likely to have allergic mothers (29, 30, 32, 33), and 2 studies reported oversampling of allergic mothers by study design (26, 30). In addition, Kiefte-de Jong et al (22) reported that mothers who used supplements were significantly different from mothers who did not use supplements for 12 of 16 maternal and child characteristics including the following: birth weight, parental history of atopy, age, ethnicity, education, and smoking and alcohol use during pregnancy. This degree of difference was unlikely to be amenable to statistical correction and likely indicative of residual confounding because of another unmeasured factor.

Of all of the studies in our review that examined folic acid supplementation, dietary folate intake, or blood folate concentrations in the periconceptional period or first trimester and risk of asthma or wheezing at any age, the only study with a significant positive association was the marginally elevated finding by Håberg et al (23) (aRR: 1.06; CI: 1.03, 1.10). Results from the other studies in our review did not confirm this finding (22, 24–26, 28, 29); however, none of these studies were of a size that would allow them to detect an estimate so modestly elevated.

Limitations of exposure assessment and definition

The determination of how much natural food folate and its synthetic form, folic acid (used in multivitamins, prenatal supplements, and food fortification) was consumed is difficult. In addition, concerns about recall and self-report can lead to a misclassification of exposure. It would be ideal to use a biomarker such as red blood cell or serum folate in mothers or cord blood folate to determine folate status. However, not only does this require sampling at the critical time of exposure, which precludes its use in retrospective studies and makes very large cohort studies difficult, it can also be very costly; hence, there were a limited number of studies that used this methodology. In addition, there were a number of different folate exposure measures reported (Tables 2 and 3), and it clearly is not possible to combine and interpret different exposures measured on different scales to generate a single summary estimate.

Limitations of outcome assessment and definition

There is no single diagnostic biomarker of asthma. Symptoms of asthma, such as wheezing, are common in infants and young children and, although correlated with increased risk of asthma later in life, are often transient (4–7). The single elevated increased risk association by Håberg et al (23) was limited to the maternal report of wheezing at the young ages of 6–18 mo; a substantial portion of these infants and children are unlikely to be diagnosed with asthma. This was evident in the studies reviewed in this article, in which an elevated risk of asthma or wheezing was reported at some ages but not others (24, 27). In the meta-analysis we performed in which reports of wheezing with those of asthma were combined, we showed a moderately elevated risk estimate (RR: 1.05; 95% CI: 1.02, 1.09). However, this result was driven by the very large cohort assessed in Håberg et al (23) (89% of the summary RR weight), and the removal of it from the analysis resulted in a nonsignificant finding (RR: 1.04; 95% CI: 0.92, 1.17) (Figure 3). Because of known limitations of the association of wheezing in infancy and early toddlerhood with later asthma, the meta-analysis that was limited to asthma in children >5 y of age should be considered the more-robust estimate of the association of periconceptional folic acid use and asthma risk in childhood.

Limitations of current review

Cohort studies included in the review reported a variety of exposures (eg, folate, folic acid, serum folate, red blood cell folate, and cord blood folate), exposure time periods (eg, early compared with late pregnancy), and outcomes (eg, wheezing at 6 mo to 6 y of age, asthma diagnosis at a variety of ages, allergies, IgE concentrations, and atopy), which rendered the question of the association between folic acid exposure in pregnancy and risk of childhood asthma difficult to answer. When possible, we attempted to group studies (first-trimester supplement use and asthma or wheezing at ages 1–8 y) but fully acknowledge the limitations of the summary estimate. To date, the studies could not rule out that higher dosages consumed in the third trimester are associated with increased risk in some groups. Note that Dunstan et al (30) examined cord blood samples from an Australian cohort (n = 484) and reported a U-shaped association with increased risk of sensitization for lower and higher tertiles compared with the middle tertile, and no associations were shown when the data were not analyzed in this manner. Observed increased risk of both higher and lower cord blood could be indicative of an optimal blood folate concentration, although it was impossible to determine whether the association was because of other characteristics of these groups or a product of multiple testing in a single study.

Research agenda

To fully address the question of a possible association of folic acid supplement use in pregnancy and increased risk of asthma in children suggested by the Hollingsworth et al (14) animal model, a number of additional studies are warranted. There are at least 3 main questions that remain to be answered:

Are there epigenetic changes in utero that are associated with risk of asthma?

Are these epigenetic changes associated with folate or folic acid supplementation at specific critical developmental windows during pregnancy?

Is there a dose effect of folate or folic acid (ie, a minimum, maximum, and optimal intake) on risk of asthma, wheezing, or allergy-related outcome?

The undertaking of a randomized control trial of the effects of folic acid supplementation in pregnancy on allergic outcomes in children would be ideal, although clearly ethically problematic. Women would have to be exposed to folic acid supplementation in the periconceptional period because of the known benefits for the prevention of neural tube defects. An alternative way to address the research questions proposed in this review is to use previous trials and follow up the children whose mothers participated in those randomized controlled trials in which exposure was randomly assigned. However, the original trials were not designed to address many of the questions surrounding the timing of supplement use during gestation and may not easily address those hypotheses.

To circumvent the need for a new randomized controlled tiral, existing cohorts and Mendelian random assignment could be used to look for interactions and effects with genetic variants involved in one-carbon metabolism and folic acid intake. Some studies have looked for associations of risk with 5,10-methylenetetrahydrofolate reductase (MTHFR) (35, 37), but none to date have looked for the interaction of intake and MTHFR. It is known that individuals with the MTHFR677 CC genotype have higher blood folate concentrations for a given folic acid intake (17, 38). Because genotype is assigned at birth, it can act as a surrogate of randomization. These types of studies require substantial sample sizes to achieve a reasonable statistical power. In addition, if the causal pathway is independent of the downstream metabolites of folic acid or dietary folate, this approach would be ineffective.

In conclusion, the analysis of the published literature to date is difficult to summarize because of heterogeneity in folate- and folic acid–exposure windows during pregnancy, a variety of asthma and allergy-associated outcomes, and the limited numbers of studies. Our meta-analysis does not support the association of folic acid supplement use in the first trimester with risk of asthma in childhood.