OBJECTIVE: To evaluate the prevailing mortality paradox that second-born twins are at higher risk of perinatal mortality than first-born twins.
METHODS: We used the 1995–1997 United States “matched multiple birth” data files assembled by the National Center for Health Statistics, for analysis of risk of perinatal mortality in first- and second-born twins (293,788 fetuses). Perinatal mortality was defined to include stillbirths after 20 weeks of gestation and neonatal deaths (deaths within the first 28 days). Gestational age-specific risk of perinatal mortality (per 1,000 total births), stillbirth (per 1,000 total births), and neonatal mortality (per 1,000 livebirths) by order of twin birth were based on the fetuses-at-risk approach. Associations between order of birth and mortality indices were evaluated by fitting multivariable logistic regression models based on the method of generalized estimating equations. These models were adjusted for several potential confounding factors.
RESULTS: Perinatal mortality was 37% higher in second-born (26.1 per 1,000 total births) than in first-born (20.3 per 1,000 total births) twins (adjusted relative risk [RR] 1.37; 95% confidence interval [CI] 1.32, 1.42). The increased risk of perinatal mortality in second-born twins was chiefly driven by a 2.46-fold (95% CI 2.29, 2.63) increase in the number of stillbirths. However, the risk of neonatal mortality was very similar between first- and second-born twins (RR 0.99, 95% CI 0.95, 1.04).
CONCLUSIONS: The increased risk of perinatal death in second-born twins is driven chiefly by increased rates of stillborn second twins. Thus, the increased mortality in second-born over first-born twins probably is an artifact of mortality comparisons.
LEVEL OF EVIDENCE: II-2
The increased perinatal mortality in second-born over first-born twins is probably an artifact of mortality comparisons.
From the Division of Epidemiology, School of Public Health, University of Medicine and Dentistry of New Jersey, Section of Epidemiology and Biostatistics, and Division of Maternal-Fetal Medicine Department of Obstetrics, Gynecology and Reproductive Sciences, University of Medicine and Dentistry of New Jersey, Robert Wood Johnson Medical School/Robert Wood Johnson University Hospital, New Brunswick, New Jersey.
Received May 20, 2003. Received in revised form August 23, 2003. Accepted September 11, 2003.
Presented in part at the 123rd Annual Meeting of the American Public Health Association; December 2002; Philadelphia, Pennsylvania.
C.V.A. is partially supported through a National Institutes of Health grant (R01-HD038902).
Address reprint requests to: Cande V. Ananth, PhD, MPH, Department of Obstetrics, Gynecology and Reproductive Sciences, UMDNJ-Robert Wood Johnson Medical School, 125 Paterson Street, New Brunswick, NJ 08901; e-mail:cande.ananth@UMDNJ.EDU.
The rate of twinning in the United States in recent years has risen sharply from 18.9 per 1,000 births in 1980 to 26.8 per 1,000 births in 1997—a relative increase of 42%.1,2 This increase has been largely attributed to the introduction of assisted reproduction techniques, increased use of ovulation-inducing drugs, and delayed childbearing.3 Compared with singletons, twins experience perinatal mortality rates 4 to 10 times higher.4
Perinatal mortality is strongly associated with birth weight, and in the absence of congenital malformations, the increased risks are driven chiefly by early delivery and fetal growth restriction. Several studies have shown that, compared with first-born twins, those delivered second carry increased risks of perinatal mortality.5–9 In fact, the increased perinatal mortality in second-born compared with first-born twins was evident at every 500-g birth weight category.7 These and other investigators5–9 collectively proposed several hypotheses to explain the increased mortality in second-born compared with first-born twins. These included 1) risk of oxygen deficiency in the second-born twin due to premature separation of the placenta after the delivery of the first twin; 2) reduced placental circulation; 3) increased interval between delivery of the two fetuses, and therefore, increased oxygen deficiency; 4) a tendency of macerated fetuses to be delivered after the birth of a liveborn twin; and 5) more frequent breech delivery among second-born twins.
The purpose of this study was to challenge the mortality “paradox” that second-born twins are at greater risk of dying during the perinatal period than first-born twins. We present recent data from the United States, comprising approximately 300,000 twins, and compare perinatal mortality rates in first- and second-born twins in an effort to better understand this mortality paradox.
MATERIALS AND METHODS
We performed a retrospective cohort study on twins delivered in 1995–1997 in the United States, derived from the “matched multiple birth” data.10 These data were assembled by the National Center for Health Statistics of the Centers for Disease Control and Prevention and included information from live-birth and fetal and infant death certificates in twin and higher-order pregnancies. Live-birth and fetal and infant death records for twins and higher-order multiple births were selected for matching by using a 3-stage algorithm that resulted in more than 99% of the records being matched accurately.10
We compared the risk of perinatal mortality as the primary outcome between groups of first-born and second-born twins. Perinatal mortality was defined as fetal deaths (after 20 weeks of gestation) plus neonatal deaths (within the first 28 days). In a subanalysis, we also compared the risk of fetal and neonatal death between first- and second-born twins. Mortality comparisons between first- and second-born twins were also examined within categories of birth weight (in 250-g intervals), as well as gestational age (in completed weeks). Gestational age assignment in these data files was predominantly based on the last menstrual period. However, in a small fraction of births (less than 5%), gestational age was assigned on the basis of a clinical estimate,10 also contained in the vital statistics data.
Finally, comparison of neonatal mortality rates in first- and second-born twins based on the underlying cause of death was also examined. For this analysis, the cause-of-death assignments were based on the International Classification of Diseases (ICD), ninth revision, and were categorized as congenital anomalies (ICD 740–759); maternal complications of pregnancy (ICD 761); complications of placenta/cord/membranes (ICD 762); short gestation, low birth weight (ICD 765); intrauterine hypoxia (ICD 768); respiratory distress syndrome (ICD 769); infections specific to the perinatal period (ICD 771); sudden infant death syndrome (ICD 798.0); and all other causes. Since cause of death among stillbirths was not available in the data set, we were unable to perform this analysis.
Comparison of perinatal mortality between first- and second-born twins was also performed in relation to the following variables: maternal race (white, black, or other races); infant sex (male or female); breech presentation (present or absent); method of delivery (vaginal or cesarean); and fetal distress (present or absent).
Gestational age–specific perinatal mortality for first- and second-born twins was analyzed with the fetuses-at-risk approach.11 This risk was calculated as the number of perinatal deaths at any gestation divided by the number of total births (stillbirths plus live births) at the start of that gestation,12 separately for the groups of first- and second-born twins. Similarly, risks of stillbirth and neonatal mortality were also derived.
Associations between order of birth (first- or second-born twin) and mortality indices (stillbirth and neonatal and perinatal mortality) were based on unadjusted relative risk (RR). Multivariate logistic regression models for mortality indices were fit with order of twin birth as the main exposure (with first twin as the reference) from which adjusted RRs and 95% confidence intervals (CIs) were derived. These models were adjusted for the following confounding variables: maternal age (categorized as less than 20, 20–24, 25–29, 30–35, and 35 years or more), gravidity (defined as number of pregnancies, including the present, and coded as gravida 1, gravida 2, and gravida 3, or more), maternal education (categorized as less than 12 or 12 years or more of completed schooling), maternal race/ethnicity (white, black, or other race/ethnicity), and marital status (married or single).
Because examining outcomes in twin fetuses clearly violates the statistical assumption of independence during statistical analysis (thereby resulting in biased estimates of variance parameters), we fit all logistic regression models based on generalized estimating equations with robust variance estimates.13
The data contained more than 308,000 twin births in the United States during the period 1995–1997. Excluded from the analysis were unmatched twin records (n = 3,542), gestational age less than 20 completed weeks (n = 4,694), implausible birth weight/gestational age records (n = 1,486),14 and birth weight less than 500 g (n = 4,503), leaving 293,788 twin fetuses for the analysis.
The study was approved by the ethics review committee of the Institutional Review Board of University of Medicine and Dentistry of New Jersey, Robert Wood Johnson Medical School, New Brunswick, New Jersey.
Table 1 depicts the characteristics of the twin births. First-born twins weighed, on average, 25 g more (2,398 ± 633 g) than second-born (2,374 ± 636 g) twins. Compared with first-born twins, rates of low and very low birth weight were slightly higher in second-born twins, and second-born twins also showed higher rates of factors related to complications of delivery: breech presentation, cesarean delivery, and fetal distress. The overall distribution of birth weight (not shown) and mean birth weight-for-gestational age did not differ between first- and second-born twins (Figure 1).
Table 2 compares the risk of death for different mortality indices between first- and second-born twins. The overall perinatal mortality was significantly higher in second-born twins (RR 1.37; 95% CI 1.32, 1.42), resulting in an excess of 58 (per 10,000 births) perinatal deaths among second-born over those among first-born twins. The increased risk of perinatal death in second-born twins was chiefly driven by an increase in the number of stillborn twins delivered second (RR 2.46; 95% CI 2.29, 2.63). However, risk of neonatal mortality was very similar between first- and second-born twins (RR 0.99; 95% CI 0.95, 1.04). The risk of neither infant mortality (RR 0.96; 95% CI 0.91, 1.01) nor postneonatal mortality (RR 1.02; 95% CI 0.93, 1.13) was different between first- and second-born twins (not shown).
Table 3 shows the risk of perinatal mortality of first- and second-born twins stratified by birth weight in 250-g increments. The increased risk of perinatal mortality in second-born twins occurs only among twins that weigh 1,500 g or more. However, the risk of stillbirth in second-born twins is consistently higher at all birth weight categories, whereas the risk of neonatal mortality appears equally distributed between first- and second-born twins at most birth weight categories.
The risk of perinatal mortality and its components, stillbirths and neonatal deaths, stratified by known risk factors for obstetric complications, are shown in Table 4. Although the rates of some risk factors are higher among second-than among first-born twins, the risk of perinatal mortality persists within strata of the risk factor. The risk of stillbirth for second-born twins appears to be the driving factor, regardless of obstetric complications.
Figure 2 shows the gestational age–specific risk of perinatal, stillbirth, and neonatal deaths in first- and second-born twins. For all fetuses at risk of perinatal death, the hazard began to increase again with advancing gestational age at 28 weeks and beyond, with the highest risk observed at postterm gestations. The risk of perinatal mortality diverged around 26 weeks, with second-born twins exhibiting a higher risk than first-born twins reflected by divergence in stillbirth but not neonatal death. The risk of stillbirth (middle panel) indicates that second-born twins are consistently at higher risk than first-born twins at all gestational ages. However, the risk of neonatal mortality (right panel) was fairly similar for first- and second-born twins.
The relationship between cause-specific neonatal death and twin order of birth is shown in Table 5. Although none of the cause-specific neonatal deaths were different between first and second-born twins, the three leading causes of death were respiratory distress syndrome, “other” causes, and lastly, congenital malformations.
Consistent with previous observations, our study demonstrates that second-born twins experienced a significantly higher risk of perinatal mortality than did first-born twins. This risk persisted when mortality was examined within strata of gestational age. However, further scrutiny of these findings revealed that the increased perinatal mortality in second-born twins was chiefly driven by increased stillbirths and not by neonatal deaths. One possible explanation for the excess mortality observed in second-born twins could be arrived at by comparing mean birth weight by gestational age for first- and second-born twins. If a lower birth weight distribution was discovered in second-born twins, then some factor or factors could have contributed to restricted fetal growth in utero and, thus, a higher mortality. Given that fetal growth (birth weight-for-gestational age) curves between first- and second-born twins were almost identical (Figure 1), it is unlikely that fetal growth restriction would have been a contributor to the increased perinatal mortality in second-born twins.
One plausible explanation for the increased risk of stillbirth among second-born compared with first-born twins is the fact that fetuses that die in utero are typically delivered second, regardless of delivery route (especially if they are delivered by cesarean). If so, the increased perinatal mortality rate among second-born compared with first-born twins is likely to be were largely driven by the contribution of fetal deaths in second-born twins, thus supporting our finding (Table 2). Within the population of live births, however, the pattern was reversed: second-born twins were less likely to die during the neonatal period than were first-born twins. Historically, stillbirths have been analyzed in the context of perinatal mortality presumably because asphyxia was a common cause of death, both during labor and shortly after birth, and also because classification of a fetal death or a live birth varied geographically.12 The etiologies of stillbirth and neonatal deaths are vastly different; a great majority of stillbirths occur before labor and are due to congenital abnormalities.12 It is therefore prudent that the components that constitute perinatal mortality be reported separately because the causes of death widely differ. Our data show that if we simply compare the risk of perinatal mortality in first- and second-born twins, the results can be misleading: the association between being born second and perinatal death disappears when stillbirths are excluded from the analysis. Future studies may better elucidate the contribution of delivery complications occurring before (antepartum) or during (intrapartum) labor to the higher risk of stillbirth observed in second-born twins.
Factors related to delivery complications have been seen as possibly contributing to the elevated risk of perinatal mortality observed in second-born twins.7–9 Buekens and Wilcox7 speculated that the increased perinatal mortality rates in second-born twins may have resulted from macerated fetuses occurring more often after the birth of a living twin. However, if a dead fetus is typically delivered second, then a macerated fetus would not necessarily be the result of a delivery complication. We examined factors related to delivery complications, among which an association with increased perinatal deaths in second-born twins would help explain the difference. As evidenced in Table 4, although the rates of perinatal death were greater in second-born twins in relation to fetal distress and route of delivery, the risk was significant regardless of the presence or absence of these factors. These findings reinforce our observations that the excess perinatal deaths in second-born over first-born twins are driven chiefly by excess stillbirths and not neonatal deaths. Furthermore, another study examined perinatal mortality in twins by examining the effect of fetal presentation of the first twin and the method of delivery15 and reported results similar to ours.
A recent study on twins from Scotland9 (excluding antepartum stillbirths) found that an increased risk of perinatal death in second-born twins was due to delivery-related complications, especially following vaginal delivery at term (after 36 weeks of gestation). These authors also reported that 1) discordance between first- and second-born twins differed significantly in preterm and term births and that the discrepancy in weight between twins was greater in pregnancies complicated by death of the second twin delivered at term; and 2) intrapartum anoxia caused 75% of perinatal deaths in second twins, and most of these resulted from mechanical problems after vaginal delivery of first-born twins. They suggested that planned cesarean deliveries at term would improve survival of the second-born twin.
The higher risk of perinatal death in second-born twins may have been due to the fact that second-born twins have more antepartum complications than first-born twins, such as those with discordant birth weight compared with twins of similar weights.3 Unequal size of fetuses is a frequent complication that accompanies twin gestations and is associated with increased perinatal mortality and morbidity.16,17 Victoria et al18 examined perinatal outcomes in twin pregnancies by assessing birth weight discordance, stratified by placental chorionicity, both of which contribute to birth complications. They found that the incidence of severe discordance (greater than 25%) was 3-fold higher in monochorionic than in dichorionic twins, and monochorionic twins experienced worse outcomes than dichorionic twins. When they compared severely discordant monochorionic and dichorionic twins, they observed significantly more deliveries before 36 weeks in the former group. The authors suggested the existence of an association between preterm delivery in monochorionic twins and severe discordance. Among monochorionic or dichorionic twin sets with severe discordance, they observed decreased placental weight and increased umbilical cord abnormalities as the most common placental findings. Such observations frequently accompany twins—more often so than singletons—and are associated with preterm delivery, twin-twin transfusion syndrome, and fetal growth restriction.
Another possible risk factor for perinatal mortality may be an increase in induced twinning because of the availability of such treatments as in vitro fertilization, which may present with different outcomes than natural twinning with respect to first- and second-born twins. Lambalk and colleagues19 compared obstetrical outcomes in twin pregnancies resulting from assisted reproduction with those outcomes in natural twin pregnancies. What is already known is that subfertile patients show an increased rate of preterm delivery, have children with lower birth weights, and experience a higher rate of perinatal mortality, even after adjusting for age, parity, and multiplicity.19 When perinatal mortality was assessed in this Dutch population, the number of perinatal deaths was 50% higher among the induced than among natural twins (odds ratio 1.5; 95% CI 1.0, 2.2), after adjusting for maternal age. Induced twins also had significantly lower birth weights and Apgar scores. The authors stated that the higher incidence of preterm deliveries at 28 weeks or earlier that was observed in induced twins contributed strongly to the higher mortality rate.
The limitations of the present study include, first of all, misclassification of birth order of the twins for different routes of delivery. In other words, would the birth order that was reported for a cesarean delivery be the same if the twin fetuses had been delivered vaginally? Second, the contribution of congenital malformation and chromosomal abnormality to the risk of stillbirth and neonatal death could have accounted for a small portion of the differences in mortality between first- and second-born twins. Third, assigning a cause of death based on ICD coding may be subject to errors. Fourth, the analysis does not take into account several of the known or suspected risk factors for the outcomes examined. Adverse outcomes could have also been attributed to socioeconomic status, prenatal care, and other such factors known to have an effect on a newborn’s health. Finally, whether placental chorionicity could have influenced the results remains to be elucidated.
Improvement of pregnancy outcomes in second-born twins might be possible upon further investigation of factors that were unavailable to us, such as placental abnormalities and induced twinning. Separating antepartum from intrapartum stillbirths and other factors not considered in our analysis might also provide further insights into the differences in stillbirth between first- and second-born twins. Studies assessing perinatal mortality in twin sets would contribute additional data on poorer outcomes in second-born twins. Pregnancy outcomes in second-born twins could also be improved with the availability of newer technology, such as Doppler ultrasound, which examines blood flow of the placenta, or use of serial ultrasounds to detect birth weight discordancy in earlier stages of a pregnancy.
In summary, we believe that the increased risk of perinatal death in second-born compared with first-born twins is merely an artifact of mortality comparisons. Future studies should provide separate analysis of stillbirth and neonatal mortality rates, and whenever possible, separate antepartum from intrapartum stillbirths—with our recommendations mirroring those previously proposed.12
1. Martin JA, Park NM. Trends in twin and triplet births: 1980–97. Natl Vital Stat Rep 1999;47(24):1–16.
2. Parker JD, Schoendorf KC, Kiely JL. A comparison of recent trends in infant mortality among twins and singletons. Paediatr Perinat Epidemiol 2001;15:12–18.
3. Contribution of assisted reproductive technology and ovulation-inducing drugs to triplet and higher-order multiple births–United States, 1980–1997. MMWR Morb Mortal Wkly Rep 2000;49:535–8.
4. Kiely JL. The epidemiology of perinatal mortality in multiple births. Bull N Y Acad Med 1990;66:618–37.
5. Puissant F, Leroy F. The fate of the second twin. Eur J Obstet Gynaecol Reprod Biol 1983;15:275–7.
6. Fowler MG, Kleinman JC, Kiely JL, Kessel SS. Double jeopardy: twin infant mortality in the United States, 1983 and 1984. Am J Obstet Gynecol 1991;165:15–22.
7. Buekens P, Wilcox A. Why do small twins have a lower mortality rate than small singletons? Am J Obstet Gynecol 1993;168:937–41.
8. Imaizumi Y. Perinatal mortality in twins and factors influencing mortality in Japan, 1980–98. Paediatr Perinat Epidemiol 2001;15:298–305.
9. Smith GCS, Pell JP, Dobbie R. Birth order, gestational age, and risk of delivery related perinatal death in twins: retrospective cohort study. BMJ 2002;325:1–5.
10. National Center for Health Statistics. 1995–97 Matched multiple birth data set. NCHS CD-ROM Series 21, Number 12. Hyattsville, MD: US Department of Health and Human Services, Centers for Disease Control and Prevention; July 2000.
11. Joseph K, Liu S, Demissie K, Wen SW, Platt RW, Ananth CV, et al, for the Fetal and Infant Health Study Group of the Canadian Perinatal Surveillance System. A parsimonious explanation for intersecting perinatal mortality curves: understanding the effect of plurality and of parity. BMC Pregnancy and Childbirth 2003. Available at: http://www.biomedcentral.com/1471–2393/3/3
. Retrieved October 22, 2003.
12. Kramer MS, Liu S, Luo Z, Yuan H, Platt RW, Joseph KS. Analysis of perinatal mortality and its components: time for a change? Am J Epidemiol 2002;156:493–7.
13. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73:13–22.
14. Alexander GR, Himes JH, Kaufman RB, Mor J, Kogan M. A United States national reference for fetal growth. Obstet Gynecol 1996;87:163–8.
15. Grisaru D, Fuchs S, Kupferminc MJ, Har-Toov J, Niv J, Lessing JB. Outcome of 306 twin deliveries according to first twin presentation and method of delivery. Am J Perinatol 2000;17:303–7.
16. Demissie K, Rhoads GG, Ananth CV, Alexander GR, Kramer MS, Kogan MD, Joseph KS. Trends in preterm births and neonatal mortality among blacks and whites in the US from 1989 to 1997. Am J Epidemiol 2001;154:307–15.
17. Ananth CV, Demissie K, Hanley ML. Birth weight discordancy and adverse perinatal outcomes in twin gestations: the effect of placental abruption. Am J Obstet Gynecol 2003;188:954–60.
18. Victoria A, Mora G, Arias F. Perinatal outcome, placental pathology, and severity of discordance in monochorionic and dichorionic twins. Obstet Gynecol 2001;97:310–5.
© 2004 The American College of Obstetricians and Gynecologists
19. Lambalk CB, van Hoff M. Natural versus induced twinning and pregnancy outcome: a Dutch nationwide survey of primiparous dizygotic twin deliveries. Fertil Steril 2001;75:731–6.