Wen, Shi Wu MB, PhD*†; Fung, Karen Fung Kee MD‡; Oppenheimer, Lawrence MD‡; Demissie, Kitaw MD, PhD§; Yang, Qiuying MD, PhD¶; Walker, Mark MD*†
From the *OMNI Research Group, Division of Maternal-Fetal Medicine, University of Ottawa, Faculty of Medicine, Ottawa, Canada; †Ottawa Health Research Institute, Ottawa, Canada; ‡Division of Maternal-Fetal Medicine, University of Ottawa, Faculty of Medicine, Ottawa, Canada; §Division of Epidemiology, School of Public Health, University of Medicine and Dentistry of New Jersey, Piscataway, New Jersey; and ¶McLauphlin Centre for Population Health Risk Assessment, Institute of Population Health, University of Ottawa, Ottawa, Ontario, Canada.
Received July 17, 2003. Received in revised form October 21, 2003. Accepted November 6, 2003.
Address reprint requests to: Dr. Shi Wu Wen, OMNI Research Group, Department of Obstetrics and Gynecology, The Ottawa Hospital, General Campus, 501 Smyth Road, Box 241, Ottawa, Canada, K1H 8L6; e-mail: firstname.lastname@example.org.
The authors thank Joyce Martin for assistance in record linkage and Yan Chen for assistance in data analysis.
Although cesarean delivery has been widely used for the delivery of twin fetuses in modern obstetrics, about half of all twins are still delivered vaginally.1–3 Second-born twins are more likely to have intrapartum complications than first-born twins.2–6 Complications of labor after vaginal delivery of the first twin may necessitate emergent cesarean delivery for the second twin.
Population-based data on cesarean delivery rates for the second twin after vaginal delivery of the first twin are sparse in the literature. Knowledge of the occurrence and factors contributing to the emergent cesarean delivery for the second twin are useful for physicians managing twin pregnancies. We therefore conducted a population-based study of twins to estimate the rate of occurrence and to identify clinical predictions of emergency cesarean delivery in the second-born twin after vaginal delivery of the first twin.
MATERIALS AND METHODS
For this population-based twin study, we used the multiple birth file created by the Centers for Disease Control and Prevention for the period between 1995 and 1997. The multiple birth file contains all twin births in the United States, with data extracted from birth certificates and discharge abstracts.7 Sets of multiples in the 1995–1997 birth file were matched by plurality, state and county of occurrence of delivery, mother's date of birth, date of last menstrual period, number of prenatal visits, level of education, weight gain during pregnancy, and date of delivery. The matching was successful for 98% of the multiple sets.7 Available study variables in this database include sociodemographic information of the parents; maternal lifestyle factors, such as smoking during pregnancy; obstetric history; complications of the pregnancy, labor and delivery; and birth weight, gestational age, and other infant variables.
Data of the first twin were used in the creating of 2 derived variables: birth weight discordance within the same pair of twins (second twin 25% smaller or 25% larger than the first twin) and operative vaginal delivery of the first twin (yes or no). The calculated values for the derived variables were assigned to corresponding second twins, and the first twins were excluded from the analysis afterward.
Because our objective was to estimate the occurrence and clinical predictors of emergent cesarean delivery in second twins after vaginal delivery of the first twin, we excluded those subjects where both twins were delivered by cesarean delivery (either elective or emergent). Second-born twins were included if complete records were available for the twin sets and if the first twin was born vaginally. Second twins with fetal deaths or live births with a gestational age less than 24 completed weeks or birth weight less than 500 g were excluded.
We first described the distribution of maternal and fetal characteristics of the study population, including the occurrence of emergent cesarean delivery. We then estimated the adjusted odds ratios (ORs) for cesarean delivery by multivariate logistic regression models. Cesarean delivery was the dependent (outcome) variable. Independent variables that entered into the regression models included both clinical predictors and maternal sociodemographic factors. Clinical predictors included in the regression models were live birth parity (0, 1, and 2 or more; 0 as reference), gestational age (less than 28 weeks, 28–31 weeks, 32–35 weeks, 36 or more weeks; 36 or more weeks as reference), very low birth weight (less than 1,500 g; yes or no, no as reference), macrosomia (more than 4,000 g; yes or no, no as reference), birth weight discordance (second twin 25% smaller than first twin, second twin 25% larger than first twin, and the remaining; the remaining as reference), instrument-assisted vaginal delivery of the first twin (yes, no; no as reference), and the following medical and obstetric complications: maternal medical complications (diabetes, pregnancy-associated hypertension, eclampsia, placental abruption, or placenta previa), abnormal labor (precipitous labor, ie, less than 3 hours; prolonged labor, ie, more than 20 hours; or dysfunctional labor), breech and other malpresentations, cephalopelvic disproportion, cord prolapse, fetal distress, present or absent (absent as reference).
Maternal sociodemographic variables included in the regression models were maternal age (less than 20, 20–29, 30–34, 35 or more, 20–29 as reference), maternal race (white, nonwhite; white as reference), maternal smoking (yes, no, not available; no as reference), marital status (married, unmarried or unknown; married as reference), and prenatal care visits initiation time (first, second, third trimester, and none; first trimester as reference). The selection of clinical predictors and their categorization were determined by obstetrician coauthors after examining the availability of the clinical information in the database. Population-attributable risks (95% confidence intervals [CIs]) were calculated according to the method described by Walter8 to estimate the impact of various clinical predictors on the use of emergent cesarean delivery for the second twin. When the OR is less that 1, the appropriate parameter of interest is “population protection fraction,” which is a totally different concept. Because our main interest was the increased population-attributable risk by various clinical predicators for cesarean delivery, for variables with an OR of less than 1 we have indicated “not applicable” for population-attributable risk. For small point estimation of population-attributable risks, the lower limit of the CI could be negative. To avoid confusion, we have rounded the negative value to 0. The regression analysis was first performed in the overall study sample and then after stratifying the data into preterm (less than 36 weeks of gestation) and term (36 or more weeks of gestation) births. We also performed supplement analyses after excluding repeated cesarean deliveries.
A total of 152,233 twin pairs in the database had complete information. Twins with missing values on birth orders (n = 18,689) and those with apparent coding error (first twin cesarean delivery but second twin vaginal delivery, n = 2,480) were excluded for analysis. An additional 2,845 second twins were excluded because of fetal death, gestational age less than 24 completed weeks, or birth weight less than 500 g. Among the 128,219 liveborn second twins, 66,374 were excluded because of cesarean delivery for both first and second twins. Thus, the final analysis included 61,845 second twins where the first twin was born vaginally.
Table 1 describes the distribution of maternal and fetal characteristics of the study subjects. About 9.45% of the second twins were delivered by cesarean after vaginal delivery of the first twin (Table 1). The emergent cesarean delivery rates for the second twin were increased in 1) infants born to mothers with medical complications or cephalopelvic disproportion or breech and other malpresentations or abnormal labor or cord prolapse or in 2) infants with a gestational age of 28 to 35 weeks or very low birth weight or macrosomia or birth weight at least 25% higher than first twin or fetal distress (Table 2). However, first twin assisted by instruments, gestational age less than 28 weeks, or birth weight at least 25% lower than first twin were related to a reduced risk of cesarean delivery in the second-born twins (Table 2). Breech and other malpresentations had the highest population attributable risk (33.2%; 95% CI 31.8%, 34.6%), although its OR for this predictor (4.10, 95% CI 3.86, 4.36) was lower than some other predictors, such as fetal distress (7.40; 95% CI 6.71, 8.16), cephalopelvic disproportion (8.93; 95% CI 6.16, 12.96), and cord prolapse (10.53; 95% CI 9.01, 12.32). The effects of clinical predictors were much stronger in term births than preterm births (Tables 3 and 4). Supplement analysis after excluding cases with repeated cesarean delivery yielded similar results (data available upon request).
Our large population-based study found that among the second-born twins where the first twins were delivered vaginally, 9.45% were delivered by cesarean. This statistic provides a measure of the magnitude of the problem. Several previous studies have reported the rates of cesarean delivery for the second twin after vaginal delivery of the first twin, with a range varying from 0.33% to 26.8% (Constantine G, Redman CW. Caesarean delivery of the second twin [letter]. Lancet 1987;618–9).2,4,9–13 Differences in study years and population made it difficult to reconcile the results from various reports. Limited study sample in previous studies that varied from less than one hundred to several hundreds is another explanation. Our study, with 61,845 eligible second twins, is the largest. Moreover, because previous studies on the issue were based on births from a single or several participating hospitals, the reported figures cannot represent the occurrence of such an event in the general practice.
The large sample size in our study allowed an assessment of several important clinical predictors of cesarean delivery for the second-born twin after vaginal delivery of the first twin simultaneously and in the mean time adjusting for demographic and medical care factors.
Our study found that cesarean delivery rates were increased in infants born to mothers with medical complications or abnormal labor or in infants with gestational age of 28 to 35 weeks, very low birth weight, macrosomia, or birth weight at least 25% higher than first twin. However, first twin assisted by instruments, gestational age less than 28 weeks, or birth weight at least 25% smaller than first twin were related to reduced risk of cesarean delivery in the second twin. The most important risk factors of cesarean delivery were breech and other malpresentation (OR 4.10; 95% CI 3.86, 4.36), fetal distress (OR 7.40; 95% CI 6.71, 8.16), cephalopelvic disproportion (OR 8.93; 95% CI 6.16, 12.96), and cord prolapse (OR 10.53; 95% CI 9.01, 12.32). Because breech and other malpresentations were common events, they accounted a large proportion of emergent cesarean delivery for the second-born twins (population-attributable risk 33.2%; 95% CI 31.8%, 34.6%).
The reasons for a stronger effect of the clinical predictors on cesarean delivery in term births than preterm births are not clearly understood. Regardless of the mechanism, however, this finding suggests that physicians should be aware of the need to prepare for emergent cesarean delivery for the second twin after vaginal delivery of the first twin when the pregnancy is approaching term.
Previous studies have also identified several risk factors for cesarean delivery for the second twin, including birth weight,3,14 gestational age,14 noncephalic presentation,3,9,15 and clinical management style.2 These studies were based on small samples, and in some of these studies, those subjects with both first and second twins delivered via cesarean were not excluded. Because majority of the cesarean deliveries for the first twin are elective, with different clinical indication than emergent cesarean deliveries, an exclusion of these twins is necessary for meaningful analysis.
We caution that the risk factors examined in our study may not necessarily be causatively related to emergent cesarean delivery. For example, fetal distress has been considered as a “soft” diagnosis, and the more liberal use of this diagnosis may result in increased rate of cesarean delivery.16,17 The diagnosis for cephalopelvic disproportion, especially at preterm, could also be problematic. Some findings were opposite to what we have expected. We have hypothesized that operative vaginal delivery for the first twin would be related to an increased risk of cesarean delivery for the second twin. However, we observed the opposite in our data. We speculate that physicians may have exercised a good prediction before applying instruments for vaginal delivery for the first twin that the second twin is likely to be delivered vaginally, otherwise they may have chosen cesarean delivery before these procedures. Our study used birth certificate data, which lack clinical details and are prone to a certain degree of coding errors.18 However, coding errors are likely to have occurred in a random fashion, which would tend to attenuate the observed effects.19
Although emergent cesarean delivery for the second twin after vaginal delivery of the first twin may be necessary in some difficult cases, many of them may be unnecessary. For example, Olofsson and Rydhström13 reviewed 803 pairs of twins born between 1973 and 1982 in their regional hospital in Sweden and found that only 0.33% of the second twins were delivered via cesarean after vaginal delivery of the first twin. This rate was suddenly increased to 7% in next year, with no apparent improvement in perinatal mortality.12 They suggest that obstetricians should change their attitude from the policy of considering cesarean delivery as the optimal mode for delivery of twins and have provided criteria for the management of twin births.12 Pons et al2 compared the rates of cesarean delivery for the second twin after vaginal delivery of the first twin between 2 obstetric units in Paris with 2 different management approaches: expectant management and active management (artificial rupture of the membranes, either spontaneous or assisted by obstetrical maneuvers immediately after delivery of the first twin) and found that active approach diminishes the likelihood of cesarean delivery for the second twin without increasing the neonatal risk. However, the study by Pons et al is an observational one and is prone to various biases. Large scale, multicenter randomized controlled trial is needed to evaluate the effectiveness of the active management strategy.
1. Demissie K, Ananth CV, Martin J, Hanley ML, MacDorman MF, Rhoads GG. Fetal and neonatal mortality among twin gestations in the United States: the role of intrapair birth weight discordance. Obstet Gynecol 2002;100:474–80.
2. Pons JC, Dommergues M, Ayoubi JM, Gelebart M, Papiernik E. Delivery of the second twin: comparison of two approaches. Eur J Obstet Gynecol Reprod Biol 2002;104:32–9.
3. Zhang J, Bowes WA Jr, Grey TW, McMahon MJ. Twin delivery and neonatal and infant mortality: a population-based study. Obstet Gynecol 1996;88:593–8.
4. Roopnarinesingh AJ. Abdominal delivery of the second twin. J Obstet Gynaecol 2002;22:379–80.
5. Smith GC, Pell JP, Dobbie R. Birth order, gestational age, and risk of delivery related perinatal death in twins: retrospective cohort study. BMJ 2002;325:1004–9.
6. Simpson LI, D'Alton ME. Multiple pregnancy. In: Creasy RK, ed. Management of labor and delivery. Boston (MA): Blackwell Science; 1997. p. 395–413.
7. National Center for Health Statistics. 1995–1997 matched multiple birth data set. NCHS CD-ROM Series 21, No. 12. Hyattsville (MD): U.S. Department of Health and Human Services, Centers for Disease Control and Prevention; 2000.
8. Walter SD. The estimation and interpretation of attributable risk in health research. Biometrics 1976;32:829–49.
9. Adam C, Allen AC, Baskett TF. Twin delivery: influence of the presentation and method of delivery on the second twin. Am J Obstet Gynecol 1991;165:23–7.
10. Adinma JIB, Agbai AO. The second twin: retained and unretained. Trop Doct 1995;25:132–3.
11. Nkata M. Perinatal mortality in retained second twin. J Obstet Gynecol 2000;20:256–8.
12. Olofsson P, Rydhström H. Twin delivery: how should the second twin be delivered? Am J Obstet Gynecol 1985;153:479–81.
13. Rattan PK, Knuppel RA, O'Brien WF, Scerbo JC. Cesarean delivery of the second twin after vaginal delivery of the first twin. Am J Obstet Gynecol 1986;154:936–40.
14. Hartley RS, Emanuel I, Hitti J. Perinatal mortality and neonatal morbidity rates among twin pairs at different gestational ages: optimal delivery timing at 37 to 38 weeks’ gestation. Am J Obstet Gynecol 2001;184:451–8.
15. Caukwell S, Murphy DJ. The effect of mode of delivery and gestational age on neonatal outcome of the non-cephalic-presenting second twin. Am J Obstet Gynecol 2002;187:1356–61.
16. Anderson GM, Lomas J. Determinants of the increasing cesarean birth rate. Ontario data 1979 to 1982. N Engl J Med 1984;311:887–92.
17. Anderson GM, Lomas J. Recent trends in cesarean section rates in Ontario. CMAJ 1989;141:1049–53.
18. Huston P, Naylor CD. Health services research: reporting on studies using secondary data sources. CMAJ 1996;155:1697–702.
19. Rothman KJ, Greenland S. Modern Epidemiology. 2nd ed. Philadelphia (PA): Lippincott-Raven Publishers, 1998.