Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
There is currently intense research and public health interest in the B-vitamin folate.1 An area of particular interest is the role of folate in the etiology and prevention of neural tube defects.2,3 The combined evidence from randomized trials, intervention studies, and observational studies has convincingly demonstrated that periconceptional intake of folic acid reduces the risk of having a baby with neural tube defects.2,4–6 Therefore, health authorities in many countries advise women to increase their intake of folate before and early in pregnancy.7 Starting in 1998, several countries in North and South America have introduced programs to fortify flour with folic acid. Reductions in the prevalence of neural tube defects ranging from 19% to 49% have been reported from the United States, Canada, and Chile after the introduction of these policies.8–10
Cautious implementation of intervention at the general population level includes the monitoring of adverse effects.11 There have been several reports of an association between use of folate or multivitamin supplements and multiple births. The first report came from a randomized trial12 in Hungary on prevention of neural tube occurrence. The twin rate was 3.8% in the multivitamin group (including 0.8 mg folic acid) and 2.7% in the trace element control group. No tendency for an increase was observed in the Medical Research Council randomized trial,13 but the total number of twin pregnancies was only 12 compared with 157 in the Hungarian trial.
With the exception of an observational study in Sweden that suggested a doubling of risk for dizygotic twins, other early reports on the issue were small and provided imprecise estimates.13–16 Recently, a large nonrandomized intervention study from China17 and observational studies from California18 and Texas19 comparing twin rates before and after fortification did not find evidence for an association between folate and twinning. However, folic acid has been used by the swine industry to increase litter size.20 The issue is of considerable international public health importance, as the posited effect of folate on multiple birth rates has been used as an argument in the United Kingdom, Sweden and other countries against the introduction of programs to fortify food with folate.
Since 1998, the Medical Birth Registry of Norway has received data on use of folate and multivitamin supplements both before and during pregnancy as well as pregnancies conceived by in vitro fertilization (IVF). We exploit this opportunity to estimate the association between folate and twins precisely, and to assess confounding by IVF. We also assess the importance of underreporting of folate use and of IVF, particularly IVF procedures performed abroad that were not included in the Norwegian IVF registration system.
The Medical Birth Registry of Norway has received data on all births in Norway since 1967. In December 1998, the registry introduced a modified notification form that included 5 new questions on the use of dietary supplements, multivitamins, and folate. The forms were filled in after the birth by attending health personnel who could simply check off use of a multivitamin supplement before or during pregnancy, as well as use of a folate supplement (folic acid) before or during pregnancy. For a multiple pregnancy, the registry received one form for each fetus or child. From all fertility clinics in Norway, the Medical Birth Registry also receives separate notification of all pregnancies occurring after IVF. We conducted a retrospective, population-register based study of 176,042 women who gave birth from December 1998 through the end of 2001. We identified important underreporting of both IVF pregnancies and use of folate.
Information on IVF pregnancies was obtained by contacting fertility clinics in Denmark (Ciconia Aarhus, Maigaard Fertilitetsklinik, Ciconia København) and Sweden (Fertilitetscentrum, Göteborg, Carl von Linné Kliniken, Uppsala). We established that at least 198 pregnancies per year reported to the Medical Birth Registry were from IVF procedures performed outside Norway. We compared data from the Danish and Swedish clinics with data from the IVF registration kept by the Medical Birth Registry and from IVF pregnancies reported on the mandatory birth registration form. From the comparison we established that at least 13% of IVF pregnancies associated with clinics in the other countries were unidentified in the Medical Birth Registry. We have no estimates of underreporting of IVF pregnancies from Norwegian fertility clinics.
Estimates of underreporting of folate use were obtained by comparing the birth notification form with forms filled in by 9,407 mothers at 18–20 weeks’ gestation who participated in the Norwegian Mother and Child Cohort Study (www.fhi.no). On the basis of this information, we estimated that 45% of women who took folate before conception were registered as nonusers in the Medical Birth Registry.
We studied the association between folate use and twinning by first using standard logistic regression analyses, accounting for maternal age, parity, and use of IVF. We then extended the model to account for the assumed levels of misclassification of IVF pregnancies and folate use, by allowing for specification of the fraction of IVF pregnancies that were not identified as such and the fraction of women who used folate but were registered as nonusers. By assuming nondifferential measurement error and no replicated measurements, the extended model was obtained from a full logistic model by summing the likelihood over the unobserved true IVF and folate status of the individuals.21 Furthermore, to study the effect of folate on mono- and dizygotic twins separately, the observed outcome (singleton birth, like-sex twins or unlike-sex twins) was treated as a partial observation of the full outcome that would have been observed if zygosity was known. This information was incorporated in the likelihood model as measurement error by assuming that like-sex and unlike-sex dizygotic twins are approximately equally prevalent (Weinberg's rule). The likelihood model was developed by one of the authors (H.K.G) and estimation was done using S-Plus (Insightful Inc, Seattle, WA; version 6.1 for Windows) and R (version 1.7.0 for Windows).22 (Further details of the study methods are available with the electronic version of this article).
We studied 176,042 singleton and twin pregnancies registered in the Medical Birth Registry of Norway from December 1998 through December 2001 (Table 1). The use of IVF was reported for 2620 of these pregnancies (1.5%). Overall, preconceptional use of folate was reported by 6% of the women, but was 4 times as frequent among women who conceived by IVF (24%). Also, the overall proportion of twins was much higher after IVF pregnancies (27%) compared with pregnancies after natural conception (1.4%).
In age and parity adjusted analyses, folate and multivitamin use both before and during pregnancy were associated with multiple births (Table 1). The strongest association for twinning was with preconceptional folate use; the odds ratio was 1.59 (95% CI = 1.41–1.78). The use of IVF weakly modified and strongly confounded the estimate of the association between preconceptional folate and twinning. When the analysis was restricted to non-IVF pregnancies, the increased risk of twinning was reduced (1.13; 0.97–1.33). The association was stronger when the outcome was restricted to unlike-sex twins pairs, with an odds ratio of 1.43 (1.12–1.83).
Among non-IVF pregnancies, we also performed separate analyses by parity (no previous pregnancies versus any). The association was strongest in women who had no previous pregnancies and weak (for unlike-sex twins) or absent (for all twins) in women who had given birth previously (Table 1).
Table 2 shows that these estimates were further attenuated when we modeled the underreporting (12.7% unidentified for IVF pregnancies and 45% for folate use). In the full extended misclassification model, the OR for twins in preconceptional folate users was 1.02 (95% CI = 0.85–1.24). Separation of the folate effect on monozygotic and dizygotic twins revealed a tendency towards a positive association for dizygotic twins (1.26; 0.91–1.73) and a negative association with monozygotic twins (0.70; 0.35–1.40), although these differences could easily be caused by chance.
Weak risks for twin pregnancies were observed for use of multivitamins and folate during pregnancy. The weak association between multivitamin use during pregnancy and twins remained after adjustment for IVF and misclassification of IVF pregnancies (Table 2).
The use of IVF is a strong confounder in the estimation of an effect of preconceptional folate use on the risk of a twin pregnancy. By excluding IVF pregnancies, the increased risk of a twin pregnancy associated with reported preconceptional use of folate dropped from 59% to 13%. Accounting for misclassification of IVF pregnancies and folate use further reduced the OR estimate to 1.02. Furthermore, our analysis constitutes an example of a situation in which underreporting of a strong confounder (IVF) introduces a biased estimate of the exposure-outcome (folate-twin) association away from an OR of 1.0.23 The effect of the misclassification adjustment shown in Table 2 is the net effect of adjustment for bias in opposite directions, as adjusting for underreporting of the exposure (vitamin use) increased the risk estimate and adjusting for underreporting of the confounder (IVF) reduced the risk estimate. For preconceptional folate, the effect of adjustment for exposure misclassification was negligible. Thus the correction of the OR from 1.13 to 1.02 was due only to accounting for the 13% underreporting of IVF pregnancies.
Czeizel and coworkers12 were the first to note the association between use of multivitamin supplements including folate and multiple births, in the Hungarian randomized trial of multivitamin prevention of neural tube defects. The treatment in this trial included folic acid plus many other vitamins, and the effect could potentially be unrelated or only partly due to folic acid. Since that study was published, data from several smaller studies have been inconclusive.13–15 However, recent ecologic analyses from California18 and Texas19 could not detect increased rates of multiple births after the introduction of folate fortification.
The only 2 large studies on this issue have reported conflicting results.16,17 Investigators from Sweden found that use of folate increased the risk of twins by 45% and that the association was even stronger for unlike-sex twins.16 These risk estimates were of similar magnitude to those we observed in the Norwegian data before accounting for IVF. Thus, the difference in results between these 2 Scandinavian studies may be explained by residual confounding by IVF in the Swedish study. In that study, the use of IVF and other reproductive technologies that increase the risk of multiple births were accounted for only indirectly (by information given by the mothers about involuntary childlessness). Investigators from China observed a weak inverse association between folate use and multiple births in a nonrandomized intervention trial.17 Differences between China and Sweden in genetic background, environment, diet, prevalence of twins, and proportion of monozygotic twins may have contributed to the different results.
We considered the possibility that the dose of folic acid could explain discrepant results among studies. In Norway, 0.2- and 0.4-mg folic acid tablets have been generally available since 1998. Before that time, 4-mg folic acid tablets had been available on special order from 1 pharmacy. Multivitamin preparations contained either no folic acid or a low dose (0.1 or 0.2 mg).24 No published information is available on the relative use of the different preparations, but the recommended dose of folic acid periconceptionally has been 0.4 mg since 1998.24 In the Swedish study, most women were reported to use folic acid in a 0.4-mg preparation, and a few used high-dose preparations (4 or 5 mg).16 In the Hungarian5 and Chinese17 trials the doses were 0.8 mg and 0.4 mg of folic acid, respectively. Thus, it does not seem that high versus low doses of folic acid can explain the heterogeneous results. The fact that studies comparing twin rates before and after fortification with folic acid are negative does not rule out an effect at the recommended dose for NTD prevention of 0.4 mg or at higher doses.
The Swedish study16 reported higher risks for unlike sex twins than for like-sex twins, suggesting that folate plays a role mostly on dizygotic twinning. Our data support this finding. Our analysis suggested that effects of folate differed for monozygotic and dizygotic twins, but our numbers are presently too low to come to a firm conclusion on this issue. It is also noteworthy that the Chinese study17 reported a decreased risk for twins associated with folate use in a population with a much higher proportion of monozygotic twins than in Scandinavia. However, a likely explanation for the stronger association in unlike-sex twins is that residual confounding due to misclassification by IVF results in stronger bias for dizygotic twins as they dominate IVF twin pregnancies.
Our data comprise all births in Norway during a 3-year period and include more than 3000 twin pairs. A strength of our data is that pregnancies after IVF are reported separately to the Medical Birth Registry. A weakness is the retrospective design, as information on preconceptional vitamin use was recorded after birth.
Also, the forms were filled in by attending health personnel at more than 50 hospitals. Therefore, it is not unlikely that the use of the folic acid and multivitamin categories on the birth notification form was subject to some interreporter variation. However, we collected information on underreporting both of folate use and IVF, and we accounted for this misclassification in the analyses. Our misclassification correction rests on the assumption that measurement error was nondifferential. Therefore, the further attenuation of the folate effect on twins after correction for underreporting of IVF, as well as the minor impact on our results of the substantial underreporting of folate, should be interpreted with the uncertainty of this assumption in mind. We did not have data on ovulation stimulation, and so we could not quantify the bias introduced by this treatment. We expect that it plays a role similar to (but weaker than) IVF as a confounder in the estimation of an effect of folate on twinning. This bias is likely to be smaller after exclusion of women who gave birth for the first time because use of assisted reproductive technologies decreases with increasing parity. Our analyses confirmed this pattern.
An association between reported use of multivitamins during pregnancy and twinning remained after adjustment. If causal, this association could indicate decreased risk of spontaneous abortion in women taking vitamins in pregnancy, thus increasing the likelihood of birth of both twins. The association, however, was modest and could be attributed to confounding by indication (ie, increased use of vitamins after a twin pregnancy was recognized).
In conclusion, confounding by IVF and misclassification of IVF, strongly bias the estimate of the effect of preconceptional folate use on twinning in Norway. After accounting for these sources of bias, we find no evidence that use of folate supplements before pregnancy increases the risk for twinning.
We are grateful to the fertility clinics in Denmark and Sweden, who provided summary numbers of Norwegian women treated at the various institutions.
1. Lucock M. Is folic acid the ultimate functional food component for disease prevention? BMJ
2. Botto LD, Moore CA, Khoury MJ, Erickson JD. Neural-tube defects. N Engl J Med
3. Wald NJ. Folic acid and the prevention of neural-tube defects. N Engl J Med
4. MRC Vitamin Study Research Group. Prevention of neural tube defects: results of the Medical Research Council Vitamin Study. Lancet
5. Czeizel AE, Dudas I. Prevention of the first occurrence of neural-tube defects by periconceptional vitamin supplementation. N Engl J Med
6. Berry RJ, Li Z, Erickson JD, et al. Prevention of neural-tube defects with folic acid in China. China-U.S. Collaborative Project for Neural Tube Defect Prevention. N Engl J Med
7. Cornel MC, Erickson JD. Comparison of national policies on periconceptional use of folic acid to prevent spina bifida and anencephaly (SBA). Teratology
8. Honein MA, Paulozzi LJ, Mathews TJ, Erickson JD, Wong LY. Impact of folic acid fortification of the US food supply on the occurrence of neural tube defects. JAMA
9. Ray JG, Meier C, Vermeulen MJ, Boss S, Wyatt PR, Cole DE. Association of neural tube defects and folic acid food fortification in Canada. Lancet
10. Castilla EE, Orioli IM, Lopez-Camelo JS, Dutra Mda G, Nazer-Herrera J. Preliminary data on changes in neural tube defect prevalence rates after folic acid fortification in South America. Am J Med Genet
11. Mills JL. Fortification of foods with folic acid–how much is enough? N Engl J Med
12. Czeizel AE, Metneki J, Dudas I. Higher rate of multiple births after periconceptional vitamin supplementation. N Engl J Med
13. Mathews F, Murphy M, Wald NJ, Hackshaw A. Twinning and folic acid use. Lancet
14. Werler MM, Cragan JD, Wasserman CR, Shaw GM, Erickson JD, Mitchell AA. Multivitamin supplementation and multiple births. Am J Med Genet
15. Lumley J, Watson L, Watson M, Bower C. Modelling the potential impact of population-wide periconceptional folate/multivitamin supplementation on multiple births. BJOG
16. Ericson A, Källén B, Åberg A. Use of multivitamins and folic acid in early pregnancy and multiple births in Sweden. Twin Res
17. Li Z, Gindler J, Wang H, et al. Folic acid supplements during early pregnancy and likelihood of multiple births: a population-based cohort study. Lancet
18. Shaw GM, Carmichael SL, Nelson V, Selvin S, Schaffer DM. Food fortification with folic acid and twinning among California infants. Am J Med Genet
19. Waller DK, Tita AT, Annegers JF. Rates of twinning before and after fortification of foods in the US with folic acid, Texas, 1996 to 1998. Paediatr Perinat Epidemiol
20. National Research Council. Nutrient Requirements of Swine
revised ed. Washington, DC: National Academy Press; 1998.
21. Carroll RJ, Ruppert D, Stefanski LA. Measurement Error in Nonlinear Models. London: Chapman and Hall; 1995.
22. Ihaka R, Gentleman R. R: a language for data analysis and graphics. J Comp Graph Stat
23. Greenland S. The effect of misclassification in the presence of covariates. Am J Epidemiol
24. Vollset SE, Lande B. Knowledge and attitudes of folate, and use of dietary supplements among women of reproductive age in Norway 1998. Acta Obstet Gynecol Scand