Twinning rates declined from the 1950s through the 1970s in most industrialized countries after controlling for age and parity1 but began to increase again in the 1980s because of increased use of assisted reproductive technologies.2–4 Monozygotic twinning is considered a random event with similar rates worldwide.1,5 Differences in overall twinning rates have been largely related to differences in rates of dizygotic twinning,2,5 which are influenced by maternal age,5–7 parity,5–6 race,5–9 family history,5,10 seasonal variation,11–13 and maternal height.5–6,14–15
Increasing use of fertility drugs is considered the primary cause of the recent increase in dizygotic twinning in most developed countries. However, a recent report from a Danish population indicates that twinning is more common among obese than nonobese women. Basso et al,16 in analyzing the Danish National Birth Registry, had to make assumptions regarding fertility treatments and the zygosity of like-sex twins. Because obesity has become more prevalent among reproductive-aged women over the past several decades,17,18 this association is of considerable public health importance. Therefore, we sought to confirm this relationship in a U.S. cohort of pregnancies that occurred before the widespread use of fertility drugs and in which zygosity was determined for the vast majority of twins.
MATERIALS AND METHODS
We analyzed data from the Collaborative Perinatal Project, a prospective cohort study of pregnancy and child health that enrolled women at 12 hospitals across the United States (Baltimore, Maryland; Boston, Massachusetts; Buffalo, New York; Memphis, Tennessee; Minneapolis, Minnesota; New Orleans, Louisiana; New York, New York (2 hospitals); Philadelphia, Pennsylvania; Portland, Oregon; Providence, Rhode Island; and Richmond, Virginia).19 This analysis was determined to be exempt from the need for institutional review board approval by the Office of Human Subjects Research, National Institutes of Health. The women were enrolled, usually at their first prenatal visit, between 1959 and 1966. These years preceded the availability of assisted reproductive technologies and other artificial fertility techniques, which are prevalent in modern populations. Although clomiphene was available at the time of the Collaborative Perinatal Project, only 2 women received this drug in the month before or the month of the last menstrual period, and both had singleton births. Therefore, the Collaborative Perinatal Project serves as an appropriate data source for examining the relationship between maternal body mass index (BMI) and twinning, free from artificial influences.
Demographic, socioeconomic, medical, genetic/familial, and behavioral information on the women, including self-reported height and prepregnant weight (in inches and pounds, respectively) were collected at entry into the study. They were then converted to metric units and used to calculate BMI as weight in kilograms divided by the square of height in meters. Maternal BMI was divided into the same 4 categories used by Basso et al:16 less than 20, 20–24.99, 25–29.99, and 30 or greater. Women with a BMI less than 20 served as the reference group. To examine the effect of height independently of BMI, the same analysis was performed with quartiles of height in centimeters as the independent variable.
We used multiple logistic regression to analyze the relationship between several maternal characteristics as independent variables and the plurality of the pregnancy (twin or singleton) as the dependent variable, controlling for confounders. In addition, information on zygosity, determined by a combination of gender, placental pathology, and analysis of nine blood groups20 was available for 81% of twin records. Therefore, the outcome of twinning, was further classified as monozygous or dizygous twin pregnancy. Chi-squared test for trend21 was used to examine distributions of maternal characteristics across BMI categories. These characteristics included maternal race (black, other), maternal age (continuous in years), parity (nulliparous, multiparous), and height (in quartiles). Women could have had more than one pregnancy represented in the database. To adjust for the nonindependence of repeated pregnancies to the same woman in the Collaborative Perinatal Project, generalized estimating equations22 were used to obtain odds ratios and 95% confidence intervals through use of PROC GENMOD in SAS 8.1 (SAS Institute, Cary, NC).
There were 58,767 pregnancies in the Collaborative Perinatal Project, and plurality was known for 56,366. The main reasons for missing plurality information were early spontaneous abortion where no fetus was recovered and loss to follow-up before delivery. Height and weight were present for 52,031 of the 56,366 pregnancies with known plurality. An additional 248 pregnancies were excluded for implausible height or weight, most likely because of coding error. This resulted in a final sample size of 51,783 singleton and twin pregnancies. Characteristics of mothers by BMI category are illustrated in Table 1. Maternal age and the proportion of multiparous and black women increased with increasing BMI. Among the 51,783 included pregnancies, 561 were of twin gestation (11 per 1,000 pregnancies). Of those, 35% were monozygous, 46% dizygous, and 19% were of undetermined zygosity. Crude and adjusted odds ratios for dizygous and monozygous twinning, according to BMI group, are illustrated in Table 2. In addition, the association between BMI and twinning for all pregnancies, regardless of zygosity, was examined, and there was a significant trend for increased risk of total twinning with increasing BMI (P < .001, data not shown). The odds of monozygous twin pregnancy were not statistically significantly related to increased prepregnancy BMI, and there was no significant trend with increasing BMI (P = .52), but the odds of dizygous twinning were significantly related to increased BMI. After adjusting for maternal race, age, parity, and height, the odds of dizygous twinning were still statistically significantly elevated among women with a BMI of 30 or greater (P = .002), and the trend for increasing risk of dizygous twinning with increasing BMI was significant (P = .001).
The effect of maternal height on twinning was not as strong as that of BMI (Table 3). There was no linear trend in odds ratios with increasing maternal height for monozygous twin pregnancies (P = .25). However, women in the tallest quartile of height had a significantly increased odds ratio for dizygous twin pregnancies, although not of the same magnitude as women with BMI over 30. The trend for increased twinning with increasing height was also significant (P = .01). Adjustment for maternal age, race, parity, and BMI had little effect on the odds ratios for maternal height.
There has been a marked increase in obesity in the United States, with the proportion of women aged 20–39 years with a BMI of 30 or more increasing from 9.3% in the early 1960s17 to 29.1% in 1999–2002.18 Increased maternal weight has been associated with increased incidence of twinning in other countries: Scotland,14 France,23 Nigeria,6 and Denmark.16 This report confirms that obesity is associated with an increased risk of twinning in a diverse U.S. population predating the common use of fertility drugs.
Between 1980 and 2002, the twin birth rate increased by 65% (from 18.9 to 31.1 per 1,000 live births or from 1.9% to 3.1% of all births) in the United States.24 An estimated one third of the increase in multiple births since the late 1970s and early 1980s has been attributed to the shift in maternal age distribution; the remainder is assumed to be the result of increased use of ovulation-inducing drugs and assisted reproductive technologies.25 Unlike triplets and other higher order multiples, where 70% are attributable to the use of ovulation-inducing drugs and assisted reproduction, only 18–34% of twin births can be attributed to these factors.26
The biologic association of maternal weight with dizygotic twinning may be mediated through elevated follicle-stimulating hormone (FSH) levels. Increased FSH concentrations have been demonstrated in women with an increased rate of dizygotic twinning.27,28 This rise in FSH is consistent with the increased occurrence of dizygotic twinning with the use of fertility drugs, which mimic naturally occurring elevated gonadotropins. Spontaneous dizygotic twinning is increased in certain families and is associated with raised concentrations of FSH.29 The mean FSH levels are highest in women who have had 2 sets of twins previously, intermediate in women who have had only 1 set of twins, and lowest in non–twin-prone women who have had singletons only.6 Spontaneous dizygotic twinning has been noted to decrease with both urbanization and starvation.14–15 FSH levels also increase with age and account for part of the maternal age effect.30 However, we are unaware of any direct relationship between obesity and elevated FSH.
Twins are at increased risk for a variety of adverse pregnancy outcomes and have significantly increased perinatal morbidity and mortality compared with singletons9,31 This report confirms the association of maternal weight and dizygotic twinning in the U.S. population, in the absence of fertility drugs. The influence of maternal weight as a factor for twinning will continue to grow in importance as the percentage of obese women in the United States continues to rise.
1. Tong S, Short RV. Dizygotic twinning as a measure of human fertility. Hum Reprod 1998;13:95–8.
2. Astolfi P, Ulizzi L, Zonta LA. Changes in twinning rate: Italy 1950–1996. Hum Reprod 2003;18:207–11.
3. Westergaard T, Wohlfahrt J, Aaby P, Melbye M. Population based study of rates of multiple pregnancies in Denmark, 1980–94. BMJ 1997;314:775–9.
4. Imaizumi Y. A comparative study of zygotic twinning and triplet rates in eight countries, 1972–1999. J Biosoc Sci 2003;35:287–302.
5. MacGillivray I, Samphier M, Little J. Factors affecting twinning. In: MacGillivray I, Campbell DM, Thompson B, editors. Twinning and twins. Great Britain (UK): John Wiley & Sons; 1988. p. 67–97.
6. Nylander P. The factors that influence twinning rates. Acta Genet Med Gemellol (Roma) 1981;30:189–202.
7. Bonnelykke B. Maternal age and parity as predictors of human twinning. Acta Genet Med Gemellol (Roma) 1990;39:329–34.
8. Waller DK, Tita ATN, Annegers JF. Rates of twinning before and after fortification of foods in the US with folic acid, Texas, 1996 to 1998. Paediatr Perinat Epidemiol 2003;17:378–83.
9. Powers WF, Kiely JL. The risks confronting twins: a national perspective. Am J Obstet Gynecol 1994;170:456–61.
10. Lichtenstein P, Olausson PO, Kallen AJ. Twin births to mothers who are twins: a registry based study. BMJ 1996;312:879–81.
11. Dionne CE, Soderstrom M, Schwartz SM. Seasonal variation of twin births in Washington State. Acta Genet Med Gemellol (Roma) 1993;42:141–9.
12. Sharma K. The twinning rates and epidemiological characteristics of births in southeast Uttar Pradesh, India. Acta Genet Med Gemellol (Roma) 1997;46:47–56.
13. Picard R, Fraser D, Hagay Z, Leiberman J. Twinning in southern Israel: seasonal variation and effects of ethnicity, maternal age and parity. J Reprod Med 1990;35:163–7.
14. Campbell DM, Campbell AJ, MacGillivray I. Maternal characteristics of women having twin pregnancies. J Biosoc Sci 1974;6:463–70.
15. Corney G, Seedburgh D, Thompson B, Campbell DM, MacGillivray I, Timlin D. Multiple and singleton pregnancy: differences between mothers as well as offspring. Prog Clin Biol Res 1981;69A:107–14.
16. Basso O, Nohr EA, Christensen K, Olsen J. Risk of twinning as a function of maternal height and body mass index. JAMA 2004;291:1564–6.
17. Flegal KM, Carrollt MD, Ogden CL, Johnson CL. Prevalence and trends in obesity among US adults, 1999–2000. JAMA 2002;288:1723–7.
18. Hedley AA, Ogden CL, Johnson CL, Carroll MD, Curtin LR, Flegal KM. Prevalence of overweight and obesity among US children, adolescents, and adults, 1999–2002. JAMA 2004;291:2847–50.
19. Niswander KG, Gordon M. The women and their pregnancies. Philadelphia (PA): W. B. Saunders; 1972.
20. Myrianthopoulos NC. Congenital malformations in twins: epidemiologic survey. Birth Defects Orig Artic Ser 1975;11:1–39.
21. Armitage1 P. Statistical methods in medical research. Oxford (UK): Blackwell Scientific Publications; 1971. p. 363–5.
22. Zeger S, Liang K-Y. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986;42:121–30.
23. Hemon D, Berger C, Lazar P. The etiology of human dizygotic twinning with special reference to spontaneous abortions. Acta Genet Med Gemellol (Roma) 1979;28:253–8.
24. Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Menacker F, Munson ML. Births: final data for 2002. Natl Vital Stat Rep 2003;52(10):1–113.
25. Martin JA, Park MM. Trends in twin and triplet births: 1980–1997. Natl Vital Stat Rep 1999;47(24):1–16.
26. Wilcox LS, Kiely JL, Melvin CL, Martin MC. Assisted reproductive technologies: estimates of their contribution to multiple births and newborn hospital days in the United States. Fertil Steril 1996;65:361–6.
27. Nylander PP. Serum levels of gonadotrophins in relation to multiple pregnancy in Nigeria. J Obstet Gynaecol Br Commonw 1973;80:651–3.
28. Soma H, Takayama M, Kiyokawa T, Akaeda T, Tokoro K. Serum gonadotropin levels in Japanese women. Obstet Gynecol 1975;46:311–2.
29. Hall JG. Twinning. Lancet 2003;362:735–43.
30. Lambalk CB, De Koning CH, Braat DD. The endocrinology of dizygotic twinning in the human. Mol Cell Endocrinol 1998;145:97–102.
© 2005 The American College of Obstetricians and Gynecologists
31. Powers WF, Wampler NS. Further defining the risks confronting twins. Am J Obstet Gynecol 1996;175:1522–8.