|
|
| |||||||||||||||||||||||||||||||||||||||||||
Background Studies in animals indicate that natural forms of vitamin A are teratogenic. Synthetic retinoids chemically similar to vitamin A cause birth defects in humans; as in animals, the defects appear to affect tissues derived from the cranial neural crest.
Methods Between October 1984 and June 1987, we identified 22,748 pregnant women when they underwent screening either by measurement of maternal serum alpha-fetoprotein or by amniocentesis. Nurse interviewers obtained information on the women's diet, medications, and illnesses during the first trimester of pregnancy, as well as information on their family and medical history and exposure to environmental agents. We obtained information on the outcomes of pregnancy from the obstetricians who delivered the babies or from the women themselves. Of the 22,748 women, 339 had babies with birth defects; 121 of these babies had defects occurring in sites that originated in the cranial neural crest.
Results For defects associated with cranial-neural-crest tissue, the ratio of the prevalence among the babies born to women who consumed more than 15,000 IU of preformed vitamin A per day from food and supplements to the prevalence among the babies whose mothers consumed 5000 IU or less per day was 3.5 (95 percent confidence interval, 1.7 to 7.3). For vitamin A from supplements alone, the ratio of the prevalence among the babies born to women who consumed more than 10,000 IU per day to that among the babies whose mothers consumed 5000 IU or less per day was 4.8 (95 percent confidence interval, 2.2 to 10.5). Using a smoothed regression curve, we found an apparent threshold near 10,000 IU per day of supplemental vitamin A. The increased frequency of defects was concentrated among the babies born to women who had consumed high levels of vitamin A before the seventh week of gestation.
Conclusions High dietary intake of preformed vitamin A appears to be teratogenic. Among the babies born to women who took more than 10,000 IU of preformed vitamin A per day in the form of supplements, we estimate that about 1 infant in 57 had a malformation attributable to the supplement.
Experiments in animals have shown that retinoids (but not carotenoids) can be teratogenic.1,4,5,6 In humans, isotretinoin, a synthetic retinoid used in the treatment of severe acne, causes congenital fetal anomalies.7,8 Lammer et al. estimated that, with fetal exposure to isotretinoin, the risk of a malformation was 25 times greater than normal.8 As in the studies in animals, a specific group of malformations ("retinoic acid embryopathy"), including those of craniofacial, cardiac, thymic, and central nervous system structures, appears to be involved.
Thus, the available evidence is consistent with the existence of a common teratogenic mechanism by which natural and synthetic retinoids affect the development of cephalic neural-crest cells and their derivatives and perhaps interfere with the closure of the neural tube.8,9,10,11,12,13,14 Recent evidence indicates that the teratogenic effect of retinoids may derive from an effect on the expression of the homeobox gene Hoxb-1 that regulates axial patterning in the embryo.15,16
Data on the teratogenicity of vitamin A in humans are scant.17,18,19,20,21 Here we report on the relation between birth defects and the intake of vitamin A from food and supplements in a prospectively studied population of more than 22,000 pregnant women.
Methods
The study cohort was originally recruited to evaluate risk factors for neural-tube defects.22 Between October 1984 and June 1987, women from the practices of more than 100 participating obstetricians were identified when they either had a maternal serum alpha-fetoprotein measurement or underwent amniocentesis. The study protocol was reviewed and approved by the Boston University Medical Center's Institutional Review Board for Human Research.
Nearly all the women were enrolled between week 15 and week 20 of pregnancy. Nurse interviewers contacted the women by telephone to obtain information on diet, medications, and illnesses during the first trimester of pregnancy, as well as general information on their family and medical history and exposure to environmental and occupational agents. The interviewers called 24,559 women, of whom 23,491 gave their consent to participate. Of those interviewed, 29 women were excluded because their interviews were incomplete, 686 because they could not be located for follow-up, and 21 because information on the outcome of pregnancy was missing. These exclusions left 22,755 women with completed interviews and follow-up information.
Retinol Intake
The interviewers asked each woman detailed questions about her diet and her use of vitamin supplements. Women were asked, "In the three months prior to pregnancy, did you take a multivitamin?" Women who said yes were asked for the brand name of the vitamin and how many times it was taken each week. Each woman was then asked specifically whether she took supplements of vitamin A, vitamin C, vitamin E, nutritional yeast, folic acid, selenium, zinc, iron, or any other nutrient during the three months before she became pregnant. Then each woman was asked for similar information about the use of multivitamins and supplements during the first three months of her pregnancy, including the brand used, the week of pregnancy during which she began to take the vitamin, the frequency of use, and any changes in intake during this period. Information on the dosage of vitamin A was obtained from information on file about each multivitamin brand, or from the subject in the case of supplements of vitamin A alone. The timing of use during pregnancy was determined according to the reported date of the last menstrual period.
Some data on the multivitamin brand, the week vitamin use began, the frequency of intake, or the dosage of single-vitamin supplements were missing for 201 women. For these women, we substituted median values for the missing values. For example, if the week of first vitamin use was missing and the brand was a prenatal formulation, we used the median starting week for all users of prenatal vitamins. We also analyzed the data after excluding women for whom we had made such substitutions.
We estimated retinol intake during each of the 12 weeks since the last menstrual period for all the subjects for whom we had information on vitamin supplements, diet, and the outcome of pregnancy. For most analyses we classified retinol intake from food and supplements according to the mean amounts ingested during the four weeks of highest consumption during the first trimester.
The women were also asked about their consumption of 50 different foods, with this question: "How often, per day, per week, or per month, did you eat one serving of the following foods during the first eight weeks of your pregnancy?" Servings were defined by the interviewers for each food item (for example, one slice of cheese or half a grapefruit).
Our food-related analysis included only foods that are sources of retinol: milk, cheese, margarine, butter, eggs, mashed potatoes (which often contain butter and milk), chicken, chicken liver, beef, beef liver, processed meats, pizza, fish, and cold breakfast cereals. Since the women were asked to name up to three brands of breakfast cereal that they ate most often, we calculated an average retinol content for cereal based on the reported brands. We used the daily intake of retinol from each of the above foods to estimate the total daily intake of retinol from the diet.
For 106 women, information on some, but fewer than half, of the retinol-containing foods was missing. For these foods, we assigned values for retinol intake, using the median daily intake for all women in the study. We excluded from our analyses 6 women for whom information for half or more of the retinol-containing foods was missing and 1 woman for whom we lacked information on vitamin supplements; these additional exclusions left 22,748 women for whom we had completed interviews and usable data on retinol intake from both foods and vitamin supplements.
Outcome of Pregnancy
Information on the outcome of pregnancy was obtained from a questionnaire mailed to the obstetrician around the expected time of delivery. If the physician did not respond, the same questionnaire was mailed to the mother. Information requested on the follow-up form included the presence of any birth defects as well as other information about complications and outcome of pregnancy. Physicians supplied the information for 76.5 percent of the pregnancies; the mothers supplied the information for the remainder.
Two coders reviewed the outcome forms, independently classifying reported birth defects according to the codes of the Centers for Disease Control and Prevention manual for birth-defect classification.23 In cases of disagreement, a third coder examined the data and made a final decision about outcome codes. During the coding, all coders were unaware of the dietary information provided by the mothers. After coding, we classified each defect into one of the following categories: craniofacial defects (e.g., oral clefts and anomalies of the ears, eyes, and nose); central nervous system defects (e.g., reduction deformities of the brain, microcephaly, and hydrocephaly in the absence of spina bifida); anomalies of the thymus; heart defects; neural-tube defects (spina bifida, anencephaly, and encephalocele); musculoskeletal defects (limb-reduction deformities, clubfoot, syndactyly, polydactyly, and other bony defects of shoulder, forearm, wrist, and hand); urogenital defects (e.g., renal agenesis, congenital hydronephrosis, other defects of the kidneys, anomalies of the external genitalia, and hypospadias); defects of the digestive tract (e.g., tracheoesophageal fistula, congenital hypertrophic pyloric stenosis, and atresia or stenosis of intestines); and other defects (e.g., agenesis or hypoplasia of the lungs, single umbilical artery, anomalies of the spleen, and cystic hygroma).
In classifying babies with more than one birth defect, we counted each baby only once, using the following hierarchy, in descending order of priority: craniofacial, central nervous system, or thymic defects; heart defects; neural-tube defects; musculoskeletal defects; urogenital defects; defects of the digestive tract; and other defects. We did not code chromosomal defects or malformations stemming from genetic causes, such as Tay–Sachs disease or cystic fibrosis, nor did we include cerebral palsy, malabsorption syndrome, a limb defect reportedly caused by an amniotic band, or conditions that were listed as birth defects on the forms but were clearly not malformations.
Craniofacial, central nervous system, thymic, and heart defects arise, at least in part, from cranial-neural-crest cells. We grouped these defects as cranial-neural-crest outcomes. We considered neural-tube defects separately; we also considered musculoskeletal and urogenital defects separately, because in some reports they have been found to be related to retinoids.7 We grouped gastrointestinal and all other defects listed above into the fourth outcome category, "other defects."
Statistical Analysis
We analyzed the data first by obtaining contingency tables for the main study variables, from which we calculated the prevalence of birth defects according to the mothers' retinol-intake category, along with prevalence ratios and approximate 95 percent confidence intervals.24 We then stratified the contingency tables according to each of several possible confounding variables, which included the age, education, and race of the mother and the maternal history or family history of birth defects. We also fitted a multiple logistic-regression model to the data that controlled for the above variables as well as folate intake, alcohol consumption, genital herpes infection, treated maternal diabetes, fever (temperature, >38.3°C [101°F]) during the first trimester of pregnancy, and the use of antiseizure medication, retinoids, or exogenous hormones.24
High vitamin A intake from supplements can result from taking a multivitamin with a high vitamin A content (some contained as much as 25,000 IU), from taking more than one multivitamin pill per day with a smaller dose of vitamin A, from taking vitamin A supplements, or from some combination of these. For analyses of supplement use, we used an intake of more than 10,000 IU of retinol from supplements as the highest dose category, because many multivitamin supplements contained as much as 10,000 IU in a single pill, and only a few contained more than 10,000 IU. To obtain a clearer picture of the shape of the dose–prevalence relation, we used quadratic splines to smooth the dose–prevalence curve.25
Results
Among the 22,748 women in this analysis, 339 had babies with birth defects that met our study criteria. Of these, 121 were of cranial-neural-crest origin. The distribution of these birth defects according to major category is shown in Table 1.
|
|
When we cross-classified retinol intake from food and intake from supplements, we found that the distribution of intake from food was nearly uncorrelated with the intake from supplements (r = 0.005). We therefore proceeded to examine retinol intake from supplements and from food separately (Table 3). Few women consumed large amounts of retinol from food alone. Among those who did, there was some indication of an increase in the prevalence of birth defects for those who consumed the highest amounts of retinol, but the small numbers make the estimate imprecise. The prevalence ratio for all birth defects among babies born to women who consumed more than 10,000 IU per day from food alone, as compared with the babies whose mothers consumed 5000 IU or less per day, was 1.8 (95 percent confidence interval, 0.8 to 4.3). For defects related to the cranial neural crest, the prevalence ratio was 2.0, but this ratio is statistically unstable, since it is based on only two cases of birth defects in babies born to women in the high retinol-intake category.
|
We then evaluated the effects of a high intake of vitamin A from food and supplements after stratifying the data according to potential confounding factors. We found little confounding by any single factor. We also constructed several multiple logistic-regression models incorporating vitamin A intake from food and supplements, with additional terms for the age, education, and race of the mother, the family history of birth defects, use of folate supplements during early pregnancy, treated maternal diabetes, alcohol consumption, genital herpes infection, fever (temperature, >38.3°C during the first trimester), and use of antiseizure medication, retinoids, or exogenous hormones. The estimates of the effects of vitamin A from both food and supplements were similar to the corresponding estimates from Table 3, even when both factors were in the same model. Thus, there was little aggregate confounding from the factors mentioned.
To improve our estimate of the shape of the dose–response curve relating the intake of supplements to the occurrence of cranial-neural-crest defects, we fitted an unrestricted quadratic-spline logistic model, using the same four intake categories and the same terms for all the potential confounding variables cited above. The smoothed exposure–effect curve, shown in Figure 1, indicates an apparent threshold near 10,000 IU of vitamin A per day from supplements. The figure also shows the curve for total retinol, which rises more slowly.
|
|
Some pregnancies included in our denominator ended in early fetal death. If these early terminations of pregnancy were unequally distributed among the categories of retinol intake, it could have biased our findings. To evaluate this possibility, we reanalyzed the data after excluding these pregnancies. The results were nearly identical to the results without these exclusions. We also reanalyzed the data with all early terminations considered as adverse events and found a similar pattern of effects. When we omitted women for whom data substitutions were used to determine vitamin A intake, the results changed very little.
Discussion
Our findings indicate that vitamin A is potentially teratogenic, but these findings relate solely to preformed vitamin A and not to beta carotene, a vitamin A precursor. We did not study beta carotene specifically, but studies in animals indicate that a high intake of beta carotene is neither toxic nor teratogenic.1,3,4,5,6
A relation between high vitamin A consumption during early pregnancy and the occurrence of birth defects is consistent with the results of studies of retinol in animals4,5 and of the effects of isotretinoin in humans7,8 and with two earlier case–control studies.20,21 The strong effect of high levels of retinol intake from supplements in our study is not easily ascribed to confounding or information bias. The stronger relation seen for vitamin A taken before or during organ formation than for vitamin A taken later is also consistent with a causal effect, but not with any plausible reporting bias. The apparently weaker relation between birth defects and retinol from food, as compared with supplements, may reflect greater error in measuring vitamin A levels in food or lower bioavailability of retinol consumed during meals.
These data appear to indicate a teratogenic effect of vitamin A at levels not far above those currently recommended. In our study population, about 1.4 percent of women averaged more than 10,000 IU of vitamin A per day from supplements. Since these data were collected, some manufacturers have decreased the retinol content of multivitamins, often substituting beta carotene. These changes may have lessened the teratogenic effects of vitamin A in the population as a whole. Among women who take more than 10,000 IU of preformed vitamin A per day from supplements, we estimate that 1 of every 57 babies is born with a birth defect attributable to the high vitamin A intake of the mother.
Supported by a Public Health Service grant (NS 19561) from the National Institute of Neurological Disorders and Stroke and by a contract with F. Hoffmann–LaRoche, Ltd.
We are indebted to Dr. Peter Glasner for his advice, to the Boston Collaborative Drug Surveillance Program for its key role in the research effort, and to Dr. Quirino Orlandi for help with the birth-defect coding.
Source Information
From the Section of Preventive Medicine and Epidemiology, Evans Department of Medicine (K.J.R., L.L.M., M.R.S., U.-S.D.T.N.), and the Center for Human Genetics (A.M.), Boston University School of Medicine, Boston; and the Chair of Hygiene, Institute of Human Pathology and Social Medicine, University G. D'Annunzio, Chieti, Italy (S.M.).
Address reprint requests to Dr. Rothman at Boston University School of Medicine, B-612, 88 E. Newton St., Boston, MA 02118.
References
| |||||||||||||||||||||||||||||||||||||||||||
Related Letters:
Teratogenicity of High Vitamin A Intake
Werler M. M., Lammer E. J., Mitchell A. A., Brent R. L., Hendrickx A. G., Holmes L. B., Miller R. K., Watkins M., Moore C., Mulinare J., Challem J. J., Hunt J. R., Rothman K. J., Moore L. L., Singer M. R., Milunsky A.
Extract |
Full Text
N Engl J Med 1996;
334:1195-1197, May 2, 1996.
Correspondence
Vitamin A Supplementation for Extremely-Low-Birth-Weight Infants
Bental Y. A., Rotschild A., Cooper P. A., Boucher B. J., Tyson J.
Extract |
Full Text
N Engl J Med 1999;
341:1697-1698, Nov 25, 1999.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2008 Massachusetts Medical Society. All rights reserved. |
