Multivitamin supplementation and multiple births

Multivitamin Supplementation and Multiple Births

Martha M. Werler,1* Janet D. Cragan,2 Cathy R. Wasserman,3 Gary M. Shaw,3 J. David Erickson,2and Allen A. Mitchell1

1Slone Epidemiology Unit, Boston University School of Public Health, Brookline, Massachusetts2Division of Birth Defects and Developmental Disabilities, National Center for Environmental Health, Centers forDisease Control and Prevention, Atlanta, Georgia

3March of Dimes Birth Defects Foundation, California Birth Defects Monitoring Program, Emeryville, California

It is well established that maternal multivi-tamin supplementation reduces the risk ofneural tube defects and evidence suggeststhat it may be associated with other repro-ductive outcomes. The present study wasprompted by a report from a randomizedtrial in Hungary which showed a 40% in-crease in multiple births among periconcep-tional vitamin users. Retrospectively col-lected data on multivitamin supplementa-tion were obtained on multiple andsingleton births from three separate stud-ies: Atlanta Birth Defects Case-ControlStudy (ABDCCS) malformed and nonmal-formed infants born 1968–1980, CaliforniaBirth Defects Monitoring Program (CB-DMP) malformed and nonmalformed in-fants born 1987–1989, and Boston UniversitySlone Epidemiology Unit Birth DefectsStudy (SEU-BDS) malformed infants born1987–1994. Supplementation was dividedinto three mutually exclusive categoriesbased on timing: ‘‘periconceptional’’ use—before through at least the third month af-ter conception; ‘‘early’’ use—beginning inthe first month and continuing through atleast the third month after conception; and‘‘later’’ use—beginning in the second orthird month after conception. For pericon-ceptional use, four of five datasets showed a30 to 60% greater prevalence of supplemen-tation among mothers of multiple births. Incontrast, this pattern was not evident for‘‘early’’ and ‘‘later’’ use. Overall, the studyfindings are tentative, due to a lack of con-sistency across all five datasets and theyshould not alter recent recommendationsrelated to folate supplementation for theprevention of neural tube defects. Am. J.Med. Genet. 71:93–96, 1997.© 1997 Wiley-Liss, Inc.

KEY WORDS: pregnancy; multiple vita-mins; epidemiology

INTRODUCTION

Maternal use of multivitamins has been linked to avariety of reproductive outcomes. The most establishedis a reduction in neural tube defect (NTD) risk frompericonceptional supplementation with folic acid[MRC, 1991; Czeizel et al., 1992]. Mounting evidencesuggests multivitamin use may reduce the risk of otherbirth defects, prematurity, and low birth weight [Czei-zel, 1993; Mulinare et al., 1995; Scholl et al., 1996;Shaw et al., 1995a,b]. One recent report, based on datacollected as part of a randomized trial in Hungary[Czeizel et al., 1994], suggested a 40% increase in mul-tiple births among periconceptional users. If this rela-tion is real, it raises the question whether such supple-mentation is causally related to embryogenesis or im-proves the viability of multiple gestations.

In the present study, we examined the relation be-tween multivitamins and multiple birth, using fiveseparate datasets from three birth defect studies. Toexplore the issue of causation versus improved viabil-ity, we examined the timing of maternal multivitaminsupplementation.

METHODS

The data were obtained from the Atlanta Birth De-fects Case-Control Study (ABDCCS), California BirthDefects Monitoring Program (CBDMP), and BostonUniversity Slone Epidemiology Unit Birth DefectsStudy (SEU-BDS). The ABDCCS identified, from apopulation-based registry, infants with major birth de-fects born between 1968 and 1980 in the greater met-ropolitan Atlanta area. Mothers of infants with birthdefects and a stratified random sample of childrenwithout birth defects were interviewed by telephonetwo-and-a-half to 14 years after delivery. The CBDMPis a population-based registry of infants with majorbirth defects delivered in many counties in California.Mothers of subjects with selected birth defects (cono-

*Correspondence to: Dr. Martha M. Werler, Slone Epidemiol-ogy Unit, 1371 Beacon St., Brookline, MA 02146.

Received 23 July 1996; Accepted 12 December 1996

American Journal of Medical Genetics 71:93–96 (1997)

© 1997 Wiley-Liss, Inc.

truncal, limb deficiency, neural tube, and oral cleft de-fects) and mothers of a random sample of infants with-out birth defects from the 1987 to 1989 birth cohortswere interviewed by telephone or in person on averagethree-and-a-half years after delivery. The SEU-BDS isa hospital-based study of infants with major birth de-fects born in the greater metropolitan areas of Bostonand Philadelphia, and in southeastern Ontario. Moth-ers were interviewed in person within six months ofdelivery between 1988 and 1994. Details on each studyare published elsewhere [Mulinare et al., 1988; Shawet al., 1996; Wasserman et al., 1996; Werler et al.,1996]. Single and multiple births were eligible for in-clusion in the ABDCCS, CBDMP, and SEU-BDS; in-terview data on mothers of multiple births were en-tered only once. The CBDMP data included two moth-ers of multiple births twice, once for each of two sets ofmultiple births.

The interviews in all three studies included detailedquestions on the use of multivitamins during themonths before and during pregnancy. Multivitaminsupplementation was divided into three mutually ex-clusive categories based on timing: ‘‘periconceptional’’was defined as use before conception through at leastthe third month after conception (three months beforeconception for ABDCCS and SEU-BDS, one month be-fore for CBDMP); ‘‘early’’ was defined as use beginningany time in the first month and continuing through atleast the third month after conception; and ‘‘later’’ wasdefined as use beginning any time in the second orthird month after conception. Exposure during the‘‘periconceptional’’ period may be related to embryogen-esis and/or the viability of multiple gestations, whereasexposure during the ‘‘later’’ period can only affect vi-ability. ‘‘Early’’ exposure, like that in the ‘‘periconcep-tional’’ period, may be related to the embryogenesisand/or the viability of multiple gestations, dependingon when use began during month 1. The reference cat-egory is women who began multivitamin supplementa-tion after the third month of pregnancy or did not takemultivitamin supplements during pregnancy.

Because multiple births have a greater risk of birthdefects [Hall, 1996] and multivitamins may indepen-dently affect birth defect risks [Czeizel, 1993; Mulinareet al., 1995; Shaw et al., 1995a,b], infants were catego-rized into those with no birth defect (‘‘non-malformed’’)and those with a birth defect present in the singleton orin at least one infant of a twin or triplet set (‘‘mal-formed’’). Five separate datasets were formed accord-ing to malformation status and study: ABDCCS andCBDMP each contributed both non-malformed andmalformed datasets and SEU-BDS contributed a mal-formed dataset only. Maternal multivitamin supple-mentation was compared between single and multiplebirths within each dataset. Information on the sex ofboth twins was available for the ABDCCS and CBDMPdata; information on zygosity was available for 81% oftwin infants with birth defects in the CBDMP dataset.Mothers whose race was other than white were ex-cluded (non-malformed: ABDCCS, n 4 708; CBDMP, n4 100; malformed: ABDCCS, n 4 1076; CBDMP, n 4140; SEU-BDS, n 4 437) because both the frequency ofmultiple births [Benirschke, 1995; Hall, 1996] and

multivitamin use [Yu et al., 1996] are known to varyacross racial groups and there were too few motherswhose race was other than white to allow for meaning-ful comparisons. Fertility medication use is positivelyassociated with multiple births [Benirschke, 1995;Hall, 1996], and since such users were few in number(ABDCCS non-malformed, n 4 33; CBDMP non-malformed, n 4 2; ABDCCS malformed, n 4 65; CB-DMP malformed, n 4 11; SEU-BDS, n 4 115), theywere excluded. Nine sets of conjoined twins were alsoexcluded.

Odds ratios and their 95% confidence intervals[Woolf, 1955] were used to estimate the risk of multiplebirths for categories of multivitamin supplementation.Mantel-Haenszel-adjusted odds ratios were estimatedfor the largest datasets.

RESULTSTable I shows the distribution of maternal multivi-

tamin supplementation among mothers of multiple andsingle births within each dataset. The numbers of sub-jects in many of the exposure categories were small,resulting in imprecise odds ratio estimates; none ex-cluded the null value. For periconceptional use, oddsratios ranged from 0.9 to 1.6 and all but one of the fivedatasets suggested a slight increase in the risk of mul-tiple birth. For early use, odds ratios ranged from 0.9 to2.1, with two datasets suggesting no association andthree suggesting increased, but unstable, risks. Lateruse was not positively associated with multiple birthsin four of the five datasets. When odds ratios were ad-justed for categories of maternal age (ABDCCS andSEU-BDS datasets only), no appreciable changes in es-timates were observed.

Twin births were divided into same-sex and opposite-sex groups for the ABDCCS and CBDMP data in TableII. The numbers of exposed subjects were small in eachdataset and no odds ratio excluded 1.0. In the ABDCCSdata, maternal multivitamin supplementation, par-ticularly periconceptional use, was generally morecommon among opposite-sex twins than among single-tons or same-sex twins. In the CBDMP data, same-sextwins among malformed subjects were positively asso-ciated with maternal multivitamin supplementationfor each exposure definition. Zygosity was available on26 of the 32 malformed twins from CBDMP. Of the ninemonozygotic twins, six were born to later users, andone each was born to early users, periconceptional us-ers, and non-users. Of the 17 dizygotic twins, nine wereborn to later users, three to early users, two to pericon-ceptional users, and three to non-users.

DISCUSSIONFor periconceptional use of multivitamins, four of the

five retrospective datasets showed a 30 to 60% greaterprevalence of supplementation among mothers of mul-tiple births, consistent with the findings in the prospec-tive Hungarian study [Czeizel et al., 1994]. In contrast,supplementation that began during the second or thirdmonth after conception generally was not more preva-lent, raising the hypothesis that multivitamins mayaffect the occurrence, rather than the viability, of mul-tiple births. There was no clear pattern of multiple

94 Werler et al.

birth risk across the five datasets for use that beganwithin the first month after conception.

If multivitamin supplementation is causally relatedto multiple gestations, one might expect the effect todiffer according to zygosity. Using same- versus oppo-site-sex as a proxy for zygosity, these data showed noconsistent relationship with periconceptional use(which might suggest a causal role) or with later use

(which might suggest improved viability); however,these estimates were based on small numbers. Amongthe CBDMP subjects on whom zygosity informationwas available, the distribution of multivitamin supple-mentation did not suggest a differential effect. Czeizelet al. [1994] reported rates of MZ and DZ twins to besimilarly increased among periconceptional supple-menters.

TABLE I. Maternal Multivitamin Supplementation* and Multiple Births

None Periconceptional Early Later

Multiple/single Multiple/single Multiple/single Multiple/single

Non-malformedABDCCSa 13/808 9/356 6/266 4/663OR (95% CI)b ref. 1.6 (0.7–3.7) 1.4 (0.6–3.8) 0.4 (0.1–1.2)

CBDMPc 3/105 3/122 2/80 6/279OR (95% CI)b ref. 0.9 (0.2–3.9) 0.9 (0.2–4.9) 0.8 (0.2–2.6)

MalformedABDCCSa 40/1413 17/460 11/380 26/1128OR (95% CI)b ref. 1.3 (0.7–2.3) 1.0 (0.5–2.1) 0.8 (0.5–1.3)

CBDMPc 5/265 4/134 5/127 18/449OR (95% CI)b ref. 1.6 (0.5–5.7) 2.1 (0.6–6.9) 2.1 (0.8–5.2)

SEU-BDSd 17/571 21/494 24/483 48/1368OR (95% CI)b ref. 1.4 (0.8–2.7) 1.6 (0.9–3.1) 1.2 (0.7–2.1)

*Mothers whose use was unknown or erratic were excluded: non-malformed ABDCCS, n 4 96;CBDMP, n 4 32; malformed ABDCCS, n 4 184; CBDMP, n 4 63; SEU-BDS, n 4 296.aABDCCS, Atlanta Birth Defects Case-Control Study.bOR (95% CI), Odds Ratio (95% Confidence Interval).cCBDMP, California Birth Defects Monitoring Program, March of Dimes.dSEU-BDS, Slone Epidemiology Unit Birth Defects Study.

TABLE II. Maternal Multivitamin Supplementation Among Same-Sex and Opposite-SexTwin Births

None Periconceptional Early Later

Multiple/single Multiple/single Multiple/single Multiple/single

Non-malformedABDCCSa

Same-sex 9/808 5/356 4/266 1/663OR (95% CI)b ref. 1.3 (0.5–3.8) 1.4 (0.5–4.4) 0.1 (0–1.1)

Opposite-sex 3/808 4/356 2/266 3/663OR (95% CI)b ref. 3.0 (0.7–11.9) 2.0 (0.4–11.0) 1.2 (0.3–5.4)

CBDMPc

Same-sex 2/105 2/122 1/80 4/279OR (95% CI)b ref. 0.9 (0.1–5.1) 0.7 (0.1–6.1) 0.8 (0.1–3.2)


MalformedABDCCSa



CBDMPc


Opposite-sex 2/265 1/134 0/127 5/449OR (95% CI)b ref. 1.0 (0.2–9.1) — 1.9 (0.4–7.5)

aABDCCS, Atlanta Birth Defects Case-Control Study.bOR (95% CI), Odds Ratio (95% Confidence Interval).cCBDMP, California Birth Defects Monitoring Program, March of Dimes.

Multivitamins and Multiple Births 95

We examined multivitamin supplementation andmultiple births from three different retrospective stud-ies. We chose to present the datasets separately, ratherthan in the aggregate, because the studies differed inthe interval between pregnancy and time of interview,questionnaires, and study populations. However, theseparate datasets were small, producing impreciseodds ratios. Indeed, for many comparisons, we cannotrule out odds ratios falling below or above or close tothe null value. Despite this limitation, we think thisanalysis is strengthened by our ability to examine nar-rowly defined time periods of supplementation in rela-tion to multiple births, using data from a variety ofstudy populations and designs. Although findings werenot consistent across all five datasets for each of thethree supplementation categories, patterns were evi-dent for periconceptional and later use.

We explored the issue of cause versus viability ofmultiple gestations only with an indirect approach—examination of the timing of maternal multivitaminsupplementation among births. However, the direct ap-proach of following all conceptions would allow moreaccurate measurement of supplementation effects onmultiple gestations. Such a study would require iden-tification of multiple gestations early in the first tri-mester, particularly given that multiple births aremore frequent among fetal losses than among livebirths [Hall, 1996]. In two clinical trials of vitaminsupplementation, similar rates of fetal loss were re-ported for vitamin users and non-users: 8% vs. 9%[MRC, 1991] and 12% vs. 11% [Czeizel and Dudas,1992], respectively, but multiple gestations among fe-tal losses were unknown.

In the three studies examined here, ascertainmentinvolved the identification of infants who were subse-quently classified as singletons or one of a set of twinsor triplets. Because we ascertained births, not mothers,the likelihood of selecting a twin or triplet for study isgreater than that of a singleton. Further, birth defectsare more common among multiple births. Thus, whilethe expected prevalence of twin births is 1.25% [Hall,1996], rates in these data ranged from 1.4 to 2.5%among the non-malformed and from 2.7 to 3.6% amongthe malformed. We think such over-sampling wouldnot be related to multivitamin supplementation andwould not bias our comparisons. However, the over-sampling may serve to narrow confidence intervals.

In consideration of confounding factors, adjustmentfor categories of maternal age (in three of the datasets)did not materially alter odds ratio estimates. Since thesmall numbers of subjects precluded more refined con-trol of maternal age, some undetected confounding mayremain. We controlled for the potential confounding ef-fects of race and the use of ovulation-inducing medica-tions by restricting analyses to white mothers who re-ported no use of fertility drugs. However, the relationbetween multivitamin supplementation and multiplebirths may vary according to maternal race, use of fer-tility drugs, and age; studies among all races, fertilitydrug users, and maternal age subgroups should be ex-plored. Of course, the possibility of confounding or ef-fect modification by other factors cannot be ruled out.

The epidemiology of multiple gestations has notbeen well described [Benirschke, 1995; Hall, 1996] andrecent work suggests the embryology may be morecomplex than previously thought, particularly formonozygotic twinning [Hall, 1996]. These data gener-ally suggest a slight positive association between peri-conceptional multivitamin supplementation and mul-tiple births. The relation between multivitamin use be-gun in the first three months following conception andthe risk of multiple births was not consistent, but thedata suggest an absence of a positive association. Theseresults reinforce the need to understand more aboutthe embryology and epidemiology of multiple gesta-tions, as well as the relation between multivitaminsupplementation and multiple gestations. Until theseissues are clarified, we believe the present findings tobe tentative and they should not alter recent recom-mendations related to folate supplementation for theprevention of neural tube defects.

REFERENCES

Benirschke K (1995): The biology of the twining process: How placentationinfluences outcome. Seminars in Perinatology 19:342–350.

Czeizel AE (1993): Prevention of congenital anomalies by periconceptionalmultivitamin supplementation. BMJ 306:1645–1648.

Czeizel AE, Dudas I (1992): Prevention of the first occurrence of neuraltube defects by periconceptional vitamin supplementation. N Engl JMed 327:1832–1835.

Czeizel AE, Metneki J, Dudas I (1994): Higher rate of multiple births afterpericonceptional vitamin supplementation (Letter). N Engl J Med 330:1687–1688.

Hall JG (1996): Twins and twinning (Editorial Comment). Am J Med Genet61:202–204.

MRC Vitamin Study Research Group (1991): Prevention of neural tubedefects: Results of the Medical Research Council Vitamin Study. Lan-cet 338:131–137.

Mulinare J, Cordero JF, Erickson JD, Berry RJ (1988): Periconceptionaluse of multivitamins and the occurrence of neural tube defects. JAMA260:3141–3145.

Mulinare J, Erickson JD, James LM, Berry RJ (1995): Does periconcep-tional use of multivitamins reduce the occurrence of birth defects? (Ab-stract). Am J Epidemiol 141:S3.

Scholl TO, Hediger ML, Schall JI, Khoo C-S, Fischer RL (1996): Dietaryand serum folate: Their influence on the outcome of pregnancy. Am JClin Nutr 63:520–525.

Shaw GM, O’Malley CD, Wasserman CR, Tolarova MM, Lammer EJ(1995a): Maternal periconceptional use of multivitamins and reducedrisk for conotruncal heart defects and limb deficiencies among off-spring. Am J Med Genet 59:536–545.

Shaw GM, Lammer EJ, Wasserman CR, O’Malley CD, Tolarova MM(1995b): Risks of orofacial clefts in children born to women using mul-tivitamins containing folic acid periconceptionally. Lancet 346:393–396.

Shaw GM, Wasserman CR, Lammer EJ, O’Malley CD, Murray JC, BasartAM, Tolarova MM (1996): Orofacial clefts, parental cigarette smoking,and transforming growth factor-alpha gene variants. Am J Hum Genet58:551–561.

Wasserman CR, Shaw GM, O’Malley CD, Tolarova MM, Lammer EJ(1996): Parental cigarette smoking and risk for congenital anomalies ofthe heart, neural tube, or limb. Teratology 53:261–267.

Werler MM, Louik C, Shapiro S, Mitchell AA (1996): Prepregnant weightin relation to risk of neural tube defects. JAMA 275:1089–1092.

Woolf B (1955): On estimating the relation between blood group and dis-ease. Annals of Human Genetics 19:251–253.

Yu SM, Keppel KG, Singh GK, Kessel W (1996): Preconceptional and pre-natal multivitamin-mineral supplement use in the 1988 National Ma-ternal and Infant Health Survey. Am J Public Health 86:240–242.

96 Werler et al.

Documents

Multivitamin supplementation and multiple births