Oral contraceptives (OCs) are currently the most common reversible method of birth control used in the United States.1 Women who are unaware of their pregnancy may continue to use OCs after conception has occurred, exposing their embryo to the hormones present in these contraceptives.
A few studies have observed that maternal use of OCs was associated with an increase in risk of specific categories of birth defects. A retrospective study that included 37 cases of neural tube defects observed a 2.5-fold increase in risk.2 Two case-control studies, comprising 155 and 537 cases, observed increases in the risk of limb defects of 17-fold and 1.7-fold, respectively.3,4 Also, a case-control study of 118 cases of urinary tract anomalies observed a 4.8-fold increase in risk.5
In contrast to these findings, 5 large case-control studies (with 1091 to 20,388 cases) concluded that there was no increased risk of congenital anomalies associated with use of OCs.6–10 A meta-analysis of 12 prospective studies concluded that there was no association between maternal exposure to OCs early in pregnancy and limb defects, congenital heart defects, or total birth defects.11 The number of birth defects in the prospective studies ranged from 18 to 1779.
However, not all of these studies published odds ratios for specific categories of birth defects, and some had limited statistical power to assess the effect of exposure to OCs during the first trimester (which occurs more rarely than exposure to OCs in the 3 months prior to conception).9,10,12 Thus, we undertook the current study to assess the effect of maternal OC use among 9986 infants with major birth defects and 4000 infants without birth defects.
This study was designed to address 2 questions: (1) Is maternal use of OCs in the first trimester associated with an increased risk of structural birth defects in 32 different categories? (2) Is maternal use of OCs in the 3 months prior to conception associated with an increased risk of any of the same 32 categories of structural birth defects?
We undertook an analysis of data from the National Birth Defects Prevention Study, an ongoing, case-control study with 10 participating sites (Arkansas, Atlanta, California, Iowa, Massachusetts, New Jersey, New York, North Carolina, Texas, and Utah). Mother-infant participants who were born on or after 1 October 1997 and had an anticipated date of delivery on or before 31 December 2003, were eligible for the current study. Cases were live-born infants, fetal deaths of at least 20 weeks gestation, and pregnancy terminations of any gestational age. Cases or controls who died after birth were also eligible. New Jersey did not ascertain cases among fetal deaths or elective terminations and Massachusetts did not ascertain cases among elective terminations. Detailed study methods have been published previously.13
We excluded infants known or strongly suspected to have single-gene conditions or chromosomal abnormalities. The controls were live-born infants with no major birth defects, randomly selected from hospital or state birth-certificate records from the same geographic areas. This study was approved by the Institutional Review Boards of the participating study centers and the Centers for Disease Control and Prevention.
Clinical geneticists classified case infants as having isolated or multiple defects based on a review of information abstracted from the medical record. An “isolated birth defect” was defined as either 1 major birth defect, 2 or more major birth defects affecting only 1 organ system, or 1 major birth defect with a well-described sequence of related defects. Cases with multiple birth defects had either 2 or more major unrelated defects in different organ systems or multiple associated major defects.14 All heart defects were confirmed by echocardiography, cardiac catheterization, surgery, or autopsy. Details on the classification of heart defects in the National Birth Defects Prevention Study have been published previously.15
Renal anomalies included only cases with an anomaly in both kidneys (renal agenesis or renal hypoplasia). Cases with cystic or dysplastic kidneys were not included. Only cases of second- or third-degree hypospadias are included in the data set because first-degree hypospadias is less consistently ascertained.16 Analyses for hypospadias were conducted after limiting controls to mothers of male infants.
Maternal interviews were conducted using a standardized, computer-based interview, primarily by telephone, in English or Spanish, no earlier than 6 weeks after the infant's estimated date of delivery and no later than 24 months after delivery. During the study time period, 1997–2003, participation rates for the interview were 70% among case mothers and 66% among control mothers. Interviews were completed within an average of 11 months from the estimated date of delivery for cases, and 9 months for controls.
We excluded 9% of cases (n = 1164) and 6% of controls (n = 312) who reported exposures to other reproductive hormones including fertility drugs, Depo-provera (Pfizer, New York, NY), Norplant (discontinued in the U.S. in 2002), the birth control “patch,” the birth control “ring,” the morning-after pill and progestin in early pregnancy to prevent miscarriage.
Early evidence suggested nonhormonal methods of contraception may be associated with birth defects through exposure to nonoxynol-9 present in spermicidal foams, jellies, and sponges and in spermicides added to some brands of condoms.17 There are also case reports of associations between the use of IUDs and birth defects.18–22 The majority of epidemiologic research has not corroborated an increased risk of birth defects among spermicide and IUD users.23–28 However, to ensure that we did not have confounding from use of IUDs and spermicides, women who used one of these methods any time between the 3 months prior to conception and the 3 months following conception and did not report any use of OCs during this period were excluded (12% of cases and 14% of controls). A large number of women reported use of both OCs and one of these nonhormonal methods; women with both exposures were retained.
The timing of exposures was based on mothers' reports of the months before or after pregnancy during which they used OCs. Unintended pregnancies were identified by questions on the maternal interview and included positive responses to whether the pregnancy was mistimed (women who wanted to be pregnant later) or unwanted (women who either did not want to be pregnant or did not care whether they were pregnant). Maternal use of OCs was categorized into one of the following mutually exclusive categories: (1) no periconceptional use of OCs (referent), (2) last used OCs 2 to 3 months prior to conception, (3) last used OCs one month prior to conception, and (4) any OC use in the 3 months following conception. Oral contraceptives suppress proliferation of the endometrium, and so women who became pregnant in the first 30 days after discontinuing their OCs (third exposure group above) may have differences in the endometrial lining at the time of conception compared with women who become pregnant in the second or third menstrual cycles after discontinuing their OCs. Recent research suggests that prior to 8 weeks of gestation a majority of embryonic nutrition may derive from uterine glands in the endometrium.29 Thus we postulated that differences in the endometrium in women who had recently discontinued OCs might interfere with nutrition to the embryo, resulting in an increase in the risk of birth defects, and therefore women in the third exposure group were assessed separately. Case and control mothers for whom information on the timing of exposure to OCs was missing (0.4% in both groups) were also excluded. Following all exclusions, this analysis included 9986 cases and 4000 controls.
We examined the possibility of confounding from maternal age, race/ethnicity, years of education, parity, smoking, intake of multivitamins containing folic acid, alcohol use, pregnancy intention, body mass index, and study site. Ethnic groups were classified as non-Hispanic white, Hispanic white, non-Hispanic black, and all other.
Crude and adjusted odds ratios (ORs) were examined for the association between maternal use of OCs and 16 categories of noncardiac birth defects and 16 categories of cardiac defects using logistic regression. We also computed 95% confidence intervals (CIs) for all ORs. Initially, adjusted odds ratios were based on a “full” model that contained the 3 OC exposure categories and covariates described above. If, after elimination of a specific covariate, the odds ratio changed by 10% or more (compared with the full model), the covariate was judged to be a confounder. Final adjusted models included only those covariates that met this criterion. The assessment of confounding variables was done separately for each birth defect. Therefore, the number of covariates in the final model differs across categories of birth defects.
We carried out similar analyses limited to isolated birth defects. Numbers were too sparse to calculate separate odds ratios among infants with multiple birth defects.
We also conducted multivariable analyses including a product term for maternal smoking during early pregnancy (yes/no) to assess the possibility that maternal smoking might enhance any associations present between the use of OCs in the first 3 months of pregnancy and specific birth defects. As statistical power was limited, we conducted this analysis only for those categories of birth defects that were associated with maternal use of OCs in the first trimester of pregnancy or had 300 cases or more.
All analyses were performed using the statistical software package SPSS 15.0 (SPSS Inc., Chicago, IL).
Seven percent (280 of 4000) of mothers of control infants reported last using OCs 2–3 months prior to conception, 5% (197 of 4000) last used them in the month immediately prior to conception and 5% (179 of 4000) used them following conception (Table 1). Of mothers who used OCs following conception, 86% (154 of 179) began using them prior to conception and were exposed only in the first or second month of pregnancy (not shown).
Control mothers who reported using OCs after conception were more likely to be black, to lack a college education, have an unintended pregnancy, smoke, or report no intake of multivitamins during the periconceptional period, compared with mothers who did not use OCs or who last used them before conception (Table 1).
The degree of confounding from all covariates was modest; ie, most of the adjusted odds ratios differed from the crude odds ratios by no more than 20% and the largest difference was 40%. We present only adjusted odds ratios in Tables 2 and 3 and indicate the specific adjustments for each birth defect. For some associations there were no confounding variables; in these instances we present the crude odds ratios.
Odds ratios for each of the 32 categories of structural birth defects in the study are shown in Table 2 (16 noncardiac birth defects) and Table 3 (16 cardiac birth defects), with cases of isolated and multiple defects combined. Three odds ratios were generated for each birth defect category: (1) exposure to OCs in the first 3 months of pregnancy; (2) exposure to OCs ending in the month immediately before conception; and (3) exposure to OCs ending 2–3 months before conception.
Mothers who used OCs in the first 3 months of pregnancy had elevated odds ratios for gastroschisis (adjusted OR = 1.8 [95% CI = 1.3–2.7]; Table 2) and hypoplastic left heart syndrome (2.3 [1.3–4.3]; Table 3). Limiting this analysis to cases with isolated birth defects, the odds ratios remained elevated for gastroschisis (1.8 [1.2–2.6]) and hypoplastic left heart syndrome (1.8 [0.9–3.5]); no other increased odds ratios were present (not shown).
Mothers who last used OCs in the month immediately prior to conception had an elevated odds ratio for anencephaly (1.7 [1.0–2.9]) and decreased odds ratios for tetralogy of Fallot (0.25 [0.10–0.62]) and anorectal atresia (0.56 [0.30–1.0]). Limiting this analysis to isolated cases of birth defects, the odds ratios changed very little for anencephaly (1.9 [1.1–3.2]) and tetralogy of Fallot (0.32 [0.13–0.78]). However, the odds ratio was no longer decreased for anorectal atresia (1.00 [0.50–2.0]) and a decreased odds ratio for isolated cases of craniosynostosis emerged (0.42 [0.21–0.87]; not shown). Among mothers in this exposure group, the adjusted odds ratios were below 1.0 for 21 of the 32 categories of birth defects.
Among mothers who last used OCs in the 2–3 months prior to conception, there were no decreased odds ratios and a single increased odds ratio for transposition of the great vessels (1.8 [1.2–2.8]). This association changed very little when the analysis was restricted to isolated cases (crude OR = 1.8 [95% CI = 1.2–2.7]).
Product terms were included for 14 categories of birth defects to assess the interaction between smoking during early pregnancy (yes/no) and use of OCs in the first 3 months of pregnancy (yes/no). None of these product terms had P values of 0.15 or less.
We assessed the association between maternal exposure to OCs in the first 3 months of pregnancy, the time period during which an effect of OCs was most plausible, and 32 different birth defects. Odds ratios were elevated for gastroschisis and hypoplastic left heart syndrome.
Werler et al30 and Torfs et al31 both reported no association between maternal use of OCs during the first trimester and the risk of gastroschisis in offspring. These studies were small (76 and 110 cases, respectively) and were limited in their ability to detect an odds ratio of 2.0. We are not aware of any studies that have reported an association between the use of OCs and hypoplastic left heart syndrome. Given 32 categories of birth defects, we would expect that by chance alone, 1.6 associations (32 × 0.05) would be statistically significant. Thus, our finding of 2 increased odds ratios may easily reflect chance. It is possible; however, that one or both of these birth defects are associated with maternal use of OCs after conception.
We did not corroborate prior reports of associations between use of OCs and neural tube defects, limb defects or urinary tract malformations. Li et al5 observed an elevated odds ratio of 4.8 (95% CI = 1.6–14) for an association between use of OCs during early pregnancy and abnormalities of the urinary tract and kidneys. We found no association between the use of OCs in early pregnancy and renal agenesis or renal hypoplasia, and the upper bound of our confidence interval suggests that an odds ratio of greater than 2.9 is unlikely. However, we did not assess birth defects occurring in the remainder of the urinary tract.
Early studies32,33 raised the possibility of an association between maternal use of progestin for infertility or to prevent early miscarriage and an increased occurrence of genital defects, especially hypospadias. However, other studies34–36 found no association or concluded that the associations observed were due to the infertility or subfertility for which the progestin was prescribed. A meta-analysis37 of exposure to sex hormones in the first trimester and the occurrence of external genital malformations, including hypospadias, also yielded an overall summary odds ratio of 1.09 (95% CI 0.9–1.32), suggesting no association. Based on an earlier data release of the National Birth Defects Prevention Study Carmichael et al38 observed that women who took progestins in the periconceptional period had an elevated risk of second- or third-degree hypospadias (OR = 3.7 [95% CI = 2.3–6.0]), but that the progestin taken as a component of oral contraceptives was not associated with increased risk. Kallen et al39 observed that among mothers who reported the use of OCs in early pregnancy there was a modest increase in the risk of hypospadias (OR = 1.5 [95% CI = 0.7–3.1]) that may be explained by chance. We also observed no association between the use of OCs in early pregnancy and an increased risk of second- or third-degree hypospadias (OR = 0.72 [95% CI = 0.46–1.1]).
Bracken et al7 reported a possible increase in the risk of anomalies (in aggregate) among OC users who also smoked 21 cigarettes or more per day compared with nonsmokers who did not use OCs. In contrast, we observed no evidence of interaction between smoking during early pregnancy (heavy and light smokers combined) and use of OCs during the first trimester and hypoplastic left heart syndrome, gastroschisis, or any of the birth defects in this study with a sample size of 300 cases or more.
Contrary to our hypothesis of an increase in risk for women who became pregnant in the first month after discontinuing OCs, these women had odds ratio below 1.0 for tetralogy of Fallot and craniosynostosis (analysis restricted to isolated cases). The odds ratio for anencephaly was above 1.0. These women also had odds ratios below 1.0 for 21 of the 32 categories of birth defects. Assuming there are no true associations in this exposure group, the probability of any birth defect having an odds ratio of less than 1.0 is 50%, and the probability that 21 or more of the 32 odds ratios will be below 1.0 is 6%; Σ [32!/(k! × (32 − k)!)] × (0.532), where k = 21 through 32 = 0.06.
It is possible that mothers who become pregnant in the first menstrual cycle after discontinuation of the birth control pill have a decreased risk of some categories of birth defects, especially heart defects. As the half-life of OCs is between 11 and 15 hours,40 such recent discontinuation is unlikely to have exposed the embryo to OCs. However, these women may be more fertile than women who conceived 2 or 3 months after discontinuing oral contraceptives. A lower rate of birth defects among highly fertile women is consistent with a study that observed a higher prevalence of infants with birth defects born to women who had difficulty becoming pregnant.41 It is also possible that a less favorable endometrium immediately after discontinuing OCs leads to a higher rate of early pregnancy loss among embryos with birth defects. As this is an unanticipated finding, it should be interpreted cautiously.
Overall, we examined a total of 96 odds ratios, of which 4 were elevated and 2 were decreased. Assuming no true associations between any of the 3 maternal exposure groups and any of the 32 birth defects in this study, by chance alone we would expect 4.8 associations (96 × 0.05) with a P value of 0.05 or less. Thus, all of the associations in this study may easily be explained by chance.
The large sample size, detailed information on OC use and the well-defined procedures for case definition and clinical review in this study allowed us to address some of the methodologic problems in previous studies. Many of the previous studies used broad categories of birth defects different from the more specific categories of birth defects currently used by clinical geneticists. Some of the earlier studies did not distinguish between women who last used OCs immediately before conception and those who continued to use them after conception.6,7
Our study has several limitations. We excluded the rarer types of structural birth defects, all types of chromosomal abnormalities and monogenic disorders. Thus, we cannot address the possibility that these types of birth defects may be associated with use of OCs. Similar to previous studies, we did not have sufficient statistical power to examine different types of OCs separately. Among the women using OCs in this study, 76% reported taking combination pills (containing both estrogen and progestin), 2% reported taking progestin-only pills, 3% reported taking newer “third-generation OC,” containing the progestin desogestrel, and 18% were unable to specify the type of OC.
To be effective, OCs must be taken daily, and failure to take them for 3–4 days results in vaginal bleeding. Thus, maternal recall of the use of OCs is likely to be considerably more accurate than recall of drugs that are taken sporadically and do not result in the possibility of pregnancy upon discontinuation. Moreover, both conceiving an unintended pregnancy while taking OCs and conceiving an intended pregnancy soon after discontinuing OCs, are likely to be memorable events. Thus, we believe that most mothers who use OCs in the periconceptional period will accurately remember doing so. Although we cannot exclude the possibility of recall bias, we think it is unlikely to play a large role in our results.
Given the participation rates in this study of 70% for cases and 66% for controls, there is also a possibility of selection bias. Not realizing they were pregnant, women with unintended pregnancies may have engaged in more risky behaviors. After reflecting on these behaviors, case mothers with unintended pregnancies may have been less comfortable participating in a telephone interview compared with control mothers who had unintended pregnancies. As unintended pregnancy is associated with exposure to OCs in early pregnancy, this would bias positive associations between OCs and birth defects toward the null. We cannot exclude the possibility that this selection bias or some other type of bias affected our results.
Prior to excluding mothers who used methods of birth control other than OCs, 4% (184 of 5008) of all mothers of control infants in this study reported exposure to OCs early in pregnancy. This proportion of OC use among control mothers during early pregnancy is similar to that in previous studies, which ranged from 2% to 4%.4,5,8,31 Overall, our findings are consistent with the majority of previous studies that found no increase in the risk of most congenital defects among mothers taking OCs immediately before or after conception.2–12
Given the multiple associations that we examined, our findings of an association between maternal use of OCs in early pregnancy and hypoplastic left heart syndrome and gastroschisis are compatible with chance, although we cannot rule out the possibility of real associations.
Study Centers: Centers participating in the National Birth Defects Prevention Study: University of Arkansas for Medical Sciences, Little Rock, AR (Charlotte Hobbs, U50/CCU613236); California March of Dimes, Berkeley, CA (Gary Shaw, U50/CCU913241); University of Iowa, Iowa City, IA (Paul Romitti, U50/CCU713238); Massachusetts Department of Public Health, Boston, MA (Marlene Anderka, U50/CCU113247); New York State Department of Health, Albany, NY (Charlotte Druschel, U50/CCU223184); University of North Carolina - Chapel Hill, Raleigh, NC (Andy Olshan, Robert Meyer, U50/CCU422096); Texas Department of State Health Services, Austin, TX (Mark Canfield, Peter Langlois, U50/CCU613232); Utah Department of Health, Salt Lake City, UT (Marcia Feldkamp, U50/CCU822097). Dr. Tunu Ramadhani replicated these analyses.
1. Hatcher RA, Nelson AL. Combined hormonal contraceptive methods. In: Hatcher RA, Trussell J, Stewart F, et al, eds. Contraceptive Technology.
18th ed. New York: Ardent Median, Inc; 2004;391–460.
2. Kasan PN, Andrews J. Oral contraception and congenital abnormalities. Br J Obstet Gynaecol
3. Kricker A, Elliot JW, Forrest JM, McCredie J. Congenital limb reduction deformities and use of oral contraceptives. Am J Obstet Gynecol
4. Czeizel AE, Kodaj I. A changing pattern in the association of oral contraceptives and the different groups of congenital limb deficiencies. Contraception
5. Li DK, Daling JR, Mueller BA, Hickok DE, Fantel AG, Weiss NS. Oral contraceptive use after conception in relation to the risk of congenital urinary tract anomalies. Teratology
6. Rothman KJ, Louik C. Oral contraceptives and birth defects. N Engl J Med
7. Bracken MB, Holford TR, White C, Kelsey JL. Role of oral contraception in congenital malformations of offspring. Int J Epidemiol
8. Lammer EJ, Cordero JF. Exogenous sex hormone exposure and the risk for major malformations. JAMA
9. Queisser-Luft A, Eggers I, Stolz G, Kieninger-Baum D, Schlaefer K. Serial examination of 20,248 newborn fetuses and infants: correlations between drug exposure and major malformations. Am J Med Genet
10. Martinez-Frias MA, Rodriquez-Pinilla E, Bermejo E, Prieto L. Prenatal exposure to sex hormones: a case-control study. Teratology
11. Bracken MB. Oral contraception and congenital malformations in offspring: a review and meta-analysis of the prospective studies. Obstet Gynecol
12. Linn S, Schoenbaum SC, Monson RR, Rosner B, Stubblefield PG, Ryan KJ. Lack of association between contraceptive usage and congenital malformations in offspring. Am J Obstet Gynecol
13. Yoon PW, Rasmussen SA, Lynberg MC, et al. The National Birth Defects Prevention Study. Public Health Rep
. 2001;116(suppl 1):32–40.
14. Rasmussen SA, Olney RS, Holmes LB, Lin AE, Keppler-Noreuil KM, Moore CA. Guidelines for case classification for the National Birth Defects Prevention Study. Birth Defects Res A Clin Mol Teratol
15. Botto LD, Lin AE, Riehle-Colarusso T, Malik S, Correa A. Seeking causes: classifying and evaluating congenital heart defects in etiologic studies. Birth Defects Res A Clin Mol Teratol
16. Dolk H, Vrijheid M, Scott JE, et al. Toward the effective surveillance of hypospadias. Environ Health Perspect
17. Jick H, Walker AM, Rothman KJ, et al. Vaginal spermicides and congenital disorders. JAMA
18. Hoyme UB, Pater TC, Haberlandt WF, Lippert TH. Cooper IUD and congenital malformation? (author's transl) [in German]. Geburtshilfe Frauenheilkd
19. Leighton PC, Evans DG, Wallis SM. Letter: IUD and congenital malformation. Br Med J
20. Weissmann-Brenner A, Lerner A, Peleg D. Transverse limb reduction and intrauterine device: case report and review of the literature. Eur J Contracept Reprod Health Care
21. Barrie H. Congenital malformation associated with intrauterine contraceptive device. Br Med J
22. Csécsei K, Szeifert GT, Papp Z. Amniotic bands associated with early rupture of amnion due to an intrauterine device. Zentralbl Gynakol
23. Mills JL, Harley EE, Reed GF, Berendes HW. Are spermicides teratogenic? JAMA
24. Louik C, Mitchell AA, Werler MM, Hanson JW, Shapiro S. Maternal exposure to spermicides in relation to certain birth defects. N Engl J Med
25. Strobino B, Kline J, Warburton D. Spermicide use and pregnancy outcome. Am J Public Health
26. Bracken MB, Vita K. Frequency of non-hormonal contraception around conception and association with congenital malformations in offspring. Am J Epidemiol
27. Tatum HJ, Schmidt FH, Jain AK. Management and outcome of pregnancies associated with the copper T intrauterine contraceptive device. Am J Obstet Gynecol
28. Layde PM, Goldberg MF, Safra MJ, Oakley GP. Failed intrauterine device contraception and limb reduction deformities: a case-control study. Fertil Steril
29. Burton GJ, Watson AL, Hempstock J, Skepper JN, Jauniaux E. Uterine glands provide histiotrophic nutrition for the human fetus during the first trimester of pregnancy. J Clin Endocrinol Metab
30. Werler MM, Mitchell AA, Shapiro S. First trimester maternal medication use in relation to gastroschisis. Teratology
31. Torfs CP, Katz EA, Bateson TF, Lam PK, Curry CJ. Maternal medications and environmental exposures as risk factors for gastroschisis. Teratology
32. Sweet RA, Schrott HG, Kurland R, Culp OS. Study of the incidence of hypospadias in Rochester, Minnesota, 1940–1970, and a case-control comparison of possible etiologic factors. Mayo Clin Proc
33. Calzolari E, Contiero MR, Roncarati E, Mattiuz PL, Volpato S. Aetiological factors in hypospadias. J Med Genet
34. Polednak AP, Janerich DT. Maternal characteristics and hypospadias: a case-control study. Teratology
35. Czeizel A, Tóth J. Correlation between the birth prevalence of isolated hypospadias and parental subfertility. Teratology
36. Källén B, Castilla EE, Kringelbach M, et al. Parental fertility and infant hypospadias: an international case-control study. Teratology
37. Raman-Wilms L, Tseng AL, Wighardt S, Einarson TR, Koren G. Fetal genital effects of first-trimester sex hormone exposure: a meta-analysis. Obstet Gynecol
38. Carmichael SL, Shaw GM, Laurent C, Croughan MS, Olney RS, Lammer EJ. Maternal progestin intake and risk of hypospadias. Arch Pediatr Adolesc Med
39. Källén B, Mastroiacovo P, Lancaster PA, et al. Oral contraceptives in the etiology of isolated hypospadias. Contraception
40. Perone N. The progestins. In: Goldzieher JW, Fotherby K, eds. Pharmocology of the Contraceptive Steroids
. New York: Raven Press; 1994:5–25.
41. Zhu JL, Basso O, Obel C, Bille C, Olsen J. Infertility, infertility treatment, and congenital malformations: Danish national birth cohort. BMJ