new england

The
journal of medicine
established in 1812 october 3, 2013 vol. 369 no. 14

A Randomized Trial of Planned Cesarean

or Vaginal Delivery for Twin Pregnancy
Jon F.R. Barrett, M.B., B.Ch., M.D., Mary E. Hannah, M.D.C.M., Eileen K. Hutton, Ph.D., Andrew R. Willan, Ph.D.,
Alexander C. Allen, M.D.C.M., B. Anthony Armson, M.D., Amiram Gafni, D.Sc., K.S. Joseph, M.D., Ph.D.,
Dalah Mason, M.P.H., Arne Ohlsson, M.D., Susan Ross, Ph.D., J. Johanna Sanchez, M.I.P.H.,
and Elizabeth V. Asztalos, M.D., for the Twin Birth Study Collaborative Group*

A BS T R AC T

BACKGROUND
Twin birth is associated with a higher risk of adverse perinatal outcomes than From the Department of Obstetrics and
singleton birth. It is unclear whether planned cesarean section results in a lower Gynaecology (J.F.R.B., M.E.H.), Centre for
Mother, Infant, and Child Research, Sunny­
risk of adverse outcomes than planned vaginal delivery in twin pregnancy. brook Research Institute (D.M., J.J.S.), and
Department of Newborn and Developmen­
METHODS tal Paediatrics (E.V.A.), Sunnybrook Health
Sciences Centre, the Program in Child
We randomly assigned women between 32 weeks 0 days and 38 weeks 6 days of gesta- Health Evaluative Sciences, SickKids Re­
tion with twin pregnancy and with the first twin in the cephalic presentation to planned search Institute, Dalla Lana School of Pub­
cesarean section or planned vaginal delivery with cesarean only if indicated. Elective lic Health (A.R.W.), and the Department of
Paediatrics, Mount Sinai Hospital (A.O.),
delivery was planned between 37 weeks 5 days and 38 weeks 6 days of gestation. The University of Toronto, Toronto; the Mid­
primary outcome was a composite of fetal or neonatal death or serious neonatal mor- wifery Education Program (E.K.H.) and the
bidity, with the fetus or infant as the unit of analysis for the statistical comparison. Department of Clinical Epidemiology and
Biostatistics (A.G.), McMaster University,
Hamilton, ON; the Departments of Paedi­
RESULTS atrics (A.C.A.) and Obstetrics and Gynae­
A total of 1398 women (2795 fetuses) were randomly assigned to planned cesarean cology (B.A.A.), Izaak Walton Killam Health
Centre, Dalhousie University, Halifax, NS;
delivery and 1406 women (2812 fetuses) to planned vaginal delivery. The rate of cesar- the Department of Obstetrics and Gynae­
ean delivery was 90.7% in the planned-cesarean-delivery group and 43.8% in the cology and the School of Population and
planned-vaginal-delivery group. Women in the planned-cesarean-delivery group deliv- Public Health, University of British Colum­
bia, Vancouver (K.S.J.); and the Depart­
ered earlier than did those in the planned-vaginal-delivery group (mean number of ment of Obstetrics and Gynaecology, Uni­
days from randomization to delivery, 12.4 vs. 13.3; P = 0.04). There was no significant versity of Alberta, Edmonton (S.R.) — all
difference in the composite primary outcome between the planned-cesarean-delivery in Canada. Address reprint requests to
Dr. Barrett at Sunnybrook Health Sci­
group and the planned-vaginal-delivery group (2.2% and 1.9%, respectively; odds ratio ences Centre, M4-172 2075 Bayview Ave.,
with planned cesarean delivery, 1.16; 95% confidence interval, 0.77 to 1.74; P = 0.49). Toronto, ON M4N 3M5, Canada, or at
jon.barrett@sunnybrook.ca.
CONCLUSIONS This article was updated on November 7,
In twin pregnancy between 32 weeks 0 days and 38 weeks 6 days of gestation, with 2013, at NEJM.org.

the first twin in the cephalic presentation, planned cesarean delivery did not sig- * The members of the Twin Birth Study Col­
nificantly decrease or increase the risk of fetal or neonatal death or serious neo­ laborative Group are listed in the Supple­
mentary Appendix, available at NEJM.org.
natal morbidity, as compared with planned vaginal delivery. (Funded by the Canadian
Institutes of Health Research; ClinicalTrials.gov number, NCT00187369; Current N Engl J Med 2013;369:1295-305.
DOI: 10.1056/NEJMoa1214939
Controlled Trials number, ISRCTN74420086.) Copyright © 2013 Massachusetts Medical Society

n engl j med 369;14 nejm.org october 3, 2013 1295

The New England Journal of Medicine
Downloaded from nejm.org on August 28, 2023. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

B
ecause of assisted reproductive segment cesarean delivery), and previous participa-
technologies, twin pregnancy occurs more tion in the Twin Birth Study.
frequently now than in the past, and it
complicates 2 to 3% of all births.1,2 Twins are at STUDY OVERSIGHT
higher risk for an adverse perinatal outcome The research ethics committee at each participating
than singletons.3,4 Planned cesarean section, as center approved the study protocol, which is avail-
compared with planned vaginal delivery, may re- able with the full text of this article at NEJM.org.
duce this risk.5 Although a small, randomized, The first, second, and last authors take responsi-
controlled trial did not show better perinatal out- bility for the accuracy and completeness of the
comes with planned cesarean section than with reported data and for the fidelity of the report to
planned vaginal delivery,6 several cohort studies the study protocol. All the women provided writ-
have shown a reduced risk of adverse perinatal ten informed consent before being enrolled.
outcomes for both twins, or for the second twin,
when twins at or near term were delivered by TREATMENT PROTOCOL
means of elective cesarean section.7-10 Despite Women were randomly assigned to planned ce-
the lack of evidence to support a policy of planned sarean section or planned vaginal delivery. Ran-
A Quick Take
cesarean section for twins at or near term, the domization was centrally controlled at the Centre
animation is
available at rates of elective cesarean section for twins have for Mother, Infant, and Child Research at Sun-
NEJM.org increased in North America and worldwide.11,12 nybrook Health Sciences Centre in Toronto with
We conducted the Twin Birth Study to com- the use of a computerized randomization pro-
pare the risk of fetal or neonatal death or serious gram stratified according to parity (0 vs. ≥1) and
neonatal morbidity with two delivery strategies gestational age (32 weeks 0 days to 33 weeks
— planned cesarean delivery or planned vaginal 6 days, 34 weeks 0 days to 36 weeks 6 days, or
delivery with cesarean delivery only if indicated 37 weeks 0 days to 38 weeks 6 days), with the use
— for twin pregnancies between 32 weeks 0 days of random block sizes.
and 38 weeks 6 days of gestation, if the leading Data were abstracted from the medical rec­
twin was in the cephalic presentation. ords at participating centers by trained study
staff and were recorded, after delivery, on stan-
ME THODS dardized data-collection forms. Participating
centers assessed fetal growth and well-being
STUDY DESIGN with the use of ultrasonography at least every
Women were eligible for the study if they had 4 weeks and with the use of nonstress or bio-
a twin pregnancy between 32 weeks 0 days and physical profile tests twice weekly if needed;
38 weeks 6 days of gestation, the first twin was were prepared to perform a cesarean section
in the cephalic presentation, and both fetuses within 30 minutes if necessary; and had anes-
were alive with an estimated weight between thetic, obstetrical, and nursing staff available in
1500 g and 4000 g, confirmed by means of ultra- the hospital at the time of planned vaginal delivery.
sonography within 7 days before randomization. Elective delivery by means of either cesarean
We enrolled women with pregnancies as early as section (for women in the planned-cesarean group)
32 weeks of gestation because many women with or labor induction (for women in the planned-
twins wish to begin planning the method of de- vaginal-delivery group) was planned between
livery at this time and because many twin births 37 weeks 5 days and 38 weeks 6 days of gestation,
are preterm. because evidence suggested that perinatal out-
Exclusion criteria were monoamniotic twins, comes would be best during this gestational-age
fetal reduction at 13 or more weeks of gestation, window.13-15 If the first twin was delivered vagi-
lethal fetal anomaly, contraindication to labor or nally in a woman in the planned-cesarean group,
vaginal delivery (e.g., fetal compromise, second a cesarean section was attempted for the second
twin substantially larger than the first twin, fetal twin, if logistically possible. For women with a
anomaly or condition that might cause mechan- planned vaginal delivery, we anticipated that more
ical problems at delivery, and previous vertical than 60% would deliver both twins vaginally.16
uterine incision or more than one previous low- The pregnancy was reassessed at the time of la-

1296 n engl j med 369;14 nejm.org october 3, 2013

The New England Journal of Medicine

Downloaded from nejm.org on August 28, 2023. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
 Planned Cesarean or Vaginal Delivery for Twins

neonatal mortality or serious neonatal morbidity.

bor, and if there was a contraindication to labor
or vaginal delivery, a cesarean delivery was un-Neonatal mortality was assessed for the period
dertaken. If labor was induced, standard meth- from 0 to 27 days after birth. Serious neonatal
ods were used, but prostaglandins were not morbidity was defined as one or more of the fol-
recommended for women who had previously lowing: birth trauma (spinal cord injury, basal or
undergone a cesarean section. depressed skull fracture, fracture of a long bone
Continuous electronic monitoring of the fetal
[humerus, radius, ulna, femur, tibia, or fibula];
heart rate was recommended during active labor. injury to a peripheral nerve [brachial plexus or
The use of oxytocin to augment labor and the phrenic or facial nerve] present at 72 hours of age
use of epidural analgesia were left to the discre-
or at discharge from the hospital; subdural or
tion of the obstetrician. After the delivery of the
intracerebral hemorrhage confirmed by mean of
first twin, the use of ultrasonography was en- ultrasonography, computed tomography [CT], or
couraged in order to check the presentation of magnetic resonance imaging [MRI]); Apgar score
the second twin. If the second twin was in the of less than 4 at 5 minutes; coma, stupor, or de-
cephalic presentation, amniotomy was delayed creased response to pain; seizures on at least two
until the fetal head was engaged and spontane- occasions before 72 hours of age; need for as-
ous vaginal delivery was anticipated, unless a sisted ventilation with the use of an endotracheal
nonreassuring fetal status required the use of tube, inserted within 72 hours after birth and re-
forceps or vacuum extraction. If the second twinmaining in place for at least 24 hours; septicemia
was not in the cephalic presentation, the obste-confirmed by means of blood culture or menin-
trician decided on the best delivery option (spon-
gitis confirmed by means of cerebrospinal fluid
taneous or assisted vaginal breech delivery, total
culture within 72 hours after birth; necrotizing
breech extraction with or without internal podalic
enterocolitis, defined as intestinal perforation,
version, external cephalic version and vaginal ce-
pneumatosis intestinalis, or air in the portal vein
phalic delivery, or intrapartum cesarean section).
diagnosed by means of surgery or radiography;
Women having a vaginal delivery were at- bronchopulmonary dysplasia, defined as the need
tended by a qualified obstetrician who was expe-for supplemental oxygen at a postnatal gesta-
rienced at vaginal twin delivery, defined a priori
tional age of 36 weeks and confirmed by means
as an obstetrician who judged himself or herselfof radiography; grade III or IV intraventricular
to be experienced at vaginal twin delivery and hemorrhage confirmed by means of ultrasonog-
whose department head agreed with this judg- raphy; or cystic periventricular leukomalacia con-
ment.17,18 Before beginning recruitment at each firmed by means of ultrasonography. Data for
center, we assigned a code number to qualified infants with the primary outcome events were
obstetricians who were considered to be experi- adjudicated, with masking of the assigned group
enced at vaginal twin delivery, and we recorded and (if possible) the method of delivery, by an
information about their qualifications and yearsadjudication committee.
of experience with vaginal twin delivery. Similar Another outcome was a composite of mater-
information was collected for other clinicians nal death or serious maternal morbidity before
who were present at delivery. 28 days post partum, defined as one or more
Infants received positive-pressure ventilation
of the following: death; hemorrhage (blood loss
with endotracheal intubation, oxygen, intravenous
≥1500 ml, need for blood transfusion, or need for
therapy, blood transfusion, surfactant, or a com-
dilation and curettage after delivery); laparotomy;
bination of these therapies if needed at the time
genital tract injury (need for hysterectomy; vulvar
of birth. Intracranial pathological findings were
or perineal hematoma requiring evacuation; broad-
assessed with the use of neonatal ultrasonography
ligament hematoma confirmed by means of ultra-
if clinically indicated. sonography, CT, or MRI; intraoperative damage to
the bladder, ureter, or bowel requiring repair; fis-
OUTCOMES tula involving the genital tract; or third-degree
For the present analysis, mothers and infants or fourth-degree perineal laceration involving the
were followed until 28 days after delivery. The anal sphincter or mucosa); thromboembolism
primary outcome was a composite of fetal or (deep-vein thrombosis, thrombophlebitis, or pul-

n engl j med 369;14 nejm.org october 3, 2013 1297

The New England Journal of Medicine
Downloaded from nejm.org on August 28, 2023. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

monary embolism) requiring anticoagulant ther- with the fetus or infant as the unit of analysis
apy; systemic infection (temperature ≥38.5°C on and generalized estimating equations to account
two or more occasions at least 24 hours apart, for the correlation between the two fetuses or
not including the first 24 hours after delivery, or infants from the same pregnancy.19,20
pneumonia confirmed by means of radiography Two models were fitted: one with treatment
or, if there was sepsis, confirmed by means of group alone and another with treatment group
blood culture); major medical life-threatening ill- and the stratification variables of parity and
ness (the acute respiratory distress syndrome, am- gestational age at randomization. A two-sided
niotic-fluid embolism, disseminated intravascu- P value of 0.05 or less was considered to indicate
lar coagulation, bowel obstruction, or paralytic statistical significance for the composite primary
ileus requiring the use of nasogastric suctioning); outcome. Since a very stringent level of signifi-
wound infection requiring prolongation of the cance (a two-sided P value of <0.002) was used
hospital stay, readmission to the hospital, or re- for the interim analyses, no adjustment for the
peated treatment as an outpatient; wound dehis- final analysis was deemed necessary. Standard
cence or breakdown; or other serious maternal logistic-regression models were used to compare
complication. Adverse events other than predefined treatment groups with respect to the maternal
measures of morbidity were to be reported to the composite outcome. Statistical significance was
independent data and safety monitoring board. set at a two-sided P value of less than 0.01 for
Secondary outcomes to be reported subsequent- the maternal composite outcome. Although not
ly included death or a poor neurodevelopmental planned a priori, two-sample t-tests were used to
outcome among the children at 2 years of cor- compare treatment-group means with respect to
rected age and problematic urinary, fecal, or flatal gestational age at delivery, time from randomiza-
incontinence among the mothers at 2 years post- tion to delivery of the first twin, and the interval
partum. Other maternal outcomes included satis- between the twin deliveries. For these analyses,
faction with the method of delivery, breast-feed- a two-sided P value of less than 0.05 was consid-
ing, quality of life, fatigue, and depression (see the ered to indicate statistical significance.
Supplementary Appendix, available at NEJM.org). Planned subgroup analyses for the primary
outcome were conducted by testing the interaction
STATISTICAL ANALYSIS term between the treatment group and the follow-
We calculated that a sample of 2800 pregnancies ing baseline variables: parity (0 vs. ≥1), gesta-
(5600 twins) was required in order to detect a re- tional age at randomization (32 weeks 0 days to
duction in the risk of the composite primary out- 33 weeks 6 days, 34 weeks 0 days to 36 weeks
come of fetal or neonatal death or serious neonatal 6 days, or 37 weeks 0 days to 38 weeks 6 days),
morbidity from 4% (on the basis of data from the maternal age (<30 years vs. ≥30 years), presenta-
Nova Scotia Atlee Perinatal Database regarding tion of the second twin (cephalic vs. noncephalic),
rates of adverse outcomes for twins with vaginal chorionicity (dichorionic vs. monochorionic), and
delivery or emergency cesarean section) to 2% with the national perinatal mortality in the mother’s
a policy of planned cesarean delivery, with 80% country of residence (<15 deaths per 1000 births,
power and a two-sided type I error of 0.05, allow- 15 to 20 deaths per 1000 births, or >20 deaths per
ing for a 10% rate of crossover between groups. 1000 births)21 (Table 1).
Two interim analyses were performed and re-
viewed by the data and safety monitoring board. R E SULT S
The first interim analysis included data from the
first 1000 women who underwent randomiza- CHARACTERISTICS OF THE PARTICIPANTS
tion, and the second included data from the first Between December 13, 2003, and April 4, 2011, we
1800 women who underwent randomization. enrolled 2804 women at 106 centers in 25 coun-
Fetal or neonatal death and maternal death tries. A total of 1398 women were randomly as-
were excluded from the analyses of neonatal and signed to planned cesarean section and 1406 to
maternal morbidity, respectively. Odds ratios planned vaginal delivery. The numbers of women
and 95% confidence intervals for the composite recruited in each country are provided in the
primary outcome with planned cesarean deliv- Supplementary Appendix. Outcome data were
ery, as compared with planned vaginal delivery, available for 1392 women (2783 fetuses or infants)
were calculated with the use of a logistic model in the planned-cesarean-delivery group and for

1298 n engl j med 369;14 nejm.org october 3, 2013

The New England Journal of Medicine

Downloaded from nejm.org on August 28, 2023. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
 Planned Cesarean or Vaginal Delivery for Twins

Table 1. Characteristics of Women and Their Pregnancies at Baseline.*

Planned Cesarean Delivery Planned Vaginal Delivery

Characteristic (N = 1393) (N = 1393)
Maternal age ≥30 yr — no. (%) 632 (45.4) 632 (45.4)
Parity ≥1 — no. (%) 857 (61.5) 856 (61.5)
Previous cesarean section — no. (%) 100 (7.2) 97 (7.0)
Gestational age at randomization
Mean — wk 34.9±1.8 34.9±1.8
<32 wk 0 days — no. (%) 0 1 (0.1)
32 wk 0 days to 33 wk 6 days — no. (%) 475 (34.1) 477 (34.2)
34 wk 0 days to 36 wk 6 days — no. (%) 679 (48.7) 665 (47.7)
37 wk 0 days to 38 wk 6 days — no. (%) 239 (17.2) 250 (17.9)
Estimated fetal weight — g†
First twin 2238±424 2238±419
Second twin 2223±413 2232±422
Chorionicity — no. (%)‡
Dichorionic and diamnionic 961 (69.0) 970 (69.6)
Monochorionic and diamnionic 334 (24.0) 326 (23.4)
Unknown 98 (7.0) 97 (7.0)
Not in labor at randomization — no./total no. (%) 1190/1392 (85.5) 1159/1393 (83.2)
Membranes ruptured at randomization — no. (%) 83 (6.0) 76 (5.5)
National perinatal mortality in mother’s country of
residence — no. (%)§
<15 deaths/1000 births 724 (52.0) 730 (52.4)
15–20 deaths/1000 births 596 (42.8) 591 (42.4)
>20 deaths/1000 births 73 (5.2) 72 (5.2)

* Plus–minus values are means ±SD. There were no significant between-group differences with respect to any of the
baseline variables.
† Data were missing for the following characteristics: estimated fetal weight of the first twin (for two women in the
planned-cesarean-delivery group and for two in the planned-vaginal-delivery group) and estimated fetal weight of the
second twin (for two women in the planned-cesarean-delivery group and for one in the planned-vaginal-delivery group).
‡ Chorionicity was determined by means of ultrasonography.
§ Countries included in the study that had fewer than 15 deaths per 1000 births were Australia, Belgium, Canada, Chile,
Croatia, Estonia, Germany, Greece, Hungary, Israel, the Netherlands, Oman, Poland, Qatar, Romania, Serbia, Spain,
the United Kingdom, the United States, and Uruguay; those with 15 to 20 deaths per 1000 births were Argentina, Brazil,
and Jamaica; and those with more than 20 deaths per 1000 births were Egypt and Jordan.

1392 women (2782 fetuses or infants) in the of women) in this group, 748 (59.2%) were per-
planned-vaginal-delivery group (Fig. 1). formed before labor. For women randomly as-
Baseline characteristics were similar in the two signed to planned vaginal delivery, 56.2% delivered
study groups (Table 1). Most women (82.4%) under­ both twins vaginally, and 4.2% had a combined
went randomization between 32 weeks 0 days and vaginal–cesarean delivery. The remaining women
36 weeks 6 days of gestation. (39.6%) had a cesarean section for both twins.
Table 2 shows the labor and delivery outcomes Of the 610 cesarean sections (43.8% of women),
for all women. Of the 1393 women randomly 412 (67.5%) were performed during labor.
assigned to planned cesarean section, 89.9% had The time from randomization to delivery
a cesarean section for the delivery of both fe- was shorter in the planned-cesarean-delivery
tuses or infants, 0.8% had a combined vaginal– group than in the planned-vaginal-delivery
cesarean delivery, and 9.3% delivered both twins group (mean days, 12.4 vs. 13.3; P = 0.04). The
vaginally. Of the 1263 cesarean sections (90.7% mean gestational age at delivery was lower in

n engl j med 369;14 nejm.org october 3, 2013 1299

The New England Journal of Medicine
Downloaded from nejm.org on August 28, 2023. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

2804 Women underwent randomization

1398 Women were assigned to planned 1406 Women were assigned to planned
cesarean delivery (2795 fetuses) vaginal delivery (2812 fetuses)

5 Women (10 fetuses) 13 Women (26 fetuses)

were lost to follow-up were lost to follow-up

1393 Women (2785 fetuses) were included 1393 Women (2786 fetuses) were included
in the study in the study

1 Woman was lost 2 Fetuses or infants 4 Fetuses or infants 1 Woman was lost
to follow-up were lost to follow-up were lost to follow-up to follow-up

1392 Women were included 2783 Fetuses or infants 2782 Fetuses or infants 1392 Women were included
in analysis of maternal were included in analysis were included in analysis in analysis of maternal
death or morbidity of primary outcome of primary outcome death or morbidity

24 Neonatal deaths or 17 Neonatal deaths or

stillbirths were excluded stillbirths were excluded

2759 Infants were included in 2765 Infants were included in

analysis of neonatal morbidity analysis of neonatal morbidity

Figure 1. Randomization, Enrollment, and Outcome Data.

Between December 13, 2003, and April 4, 2011, the study enrolled 2804 women at 106 centers in 25 countries. One woman in the
planned-cesarean-delivery group had a singleton pregnancy that had been thought to be a twin pregnancy during ultrasonographic
­examination at randomization.

the planned-cesarean-delivery group than in the ables to the model did not materially change
planned-vaginal-delivery group (P = 0.01). the result (odds ratio, 1.16; 95% CI, 0.77 to 1.74;
The characteristics of labor and delivery for P = 0.49). The only stratification variable that was
women having labor and for women having a significantly related to the primary outcome was
vaginal delivery are provided in Table S4 in the gestational age at randomization (odds ratio for
Supplementary Appendix. For 95.2% of the women 35 weeks 0 days to 36 weeks 6 days vs. 37 weeks
who were assigned to the planned-vaginal-delivery 0 days to 38 weeks 6 days of gestation, 1.83;
group and who had a vaginal delivery for the first and odds ratio for 32 weeks 0 days to 33 weeks
twin, an experienced obstetrician, according to 6 days vs. 37 weeks 0 days to 38 weeks 6 days,
our a priori definition, was present at the time of 3.36; P<0.001 for the overall comparison).
vaginal delivery. There was no significant difference between the
Table 3 shows the outcomes involving fetal planned-cesarean-delivery and planned-vaginal-
and neonatal death and serious neonatal mor- delivery groups in the frequency of the maternal
bidity. The frequency of the composite primary composite outcome (7.3% and 8.5%, respectively;
outcome did not differ significantly between P = 0.29) (Table 4). All adverse events documented
the planned-cesarean-delivery group and the during the trial were among the predefined
planned-vaginal-delivery group (2.2% and 1.9%, measures of morbidity composing the morbidity
respectively; odds ratio with planned cesarean component of the primary outcome; no other
delivery, 1.16; 95% confidence interval [CI], 0.77 adverse outcomes were reported to the data and
to 1.74; P = 0.49). Adding the stratification vari- safety monitoring board.

1300 n engl j med 369;14 nejm.org october 3, 2013

The New England Journal of Medicine

Downloaded from nejm.org on August 28, 2023. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
 Planned Cesarean or Vaginal Delivery for Twins

Table 2. Characteristics of Labor and Delivery for All Pregnancies.*

Planned Cesarean Delivery Planned Vaginal Delivery

Characteristic (N = 1393) (N = 1393)
Mode of delivery — no./total no. (%)
Cesarean for both† 1252/1392 (89.9) 551/1393 (39.6)
Vaginal and cesarean 11/1392 (0.8) 59/1393 (4.2)
Vaginal for both 129/1392 (9.3) 783/1393 (56.2)
Timing of cesarean section — no./total no. (%)
Before the onset of labor 748/1391 (53.8) 196/1391 (14.1)
During labor 514/1391 (37.0) 412/1391 (29.6)
No cesarean section 129/1391 (9.3) 783/1391 (56.3)
Presentation at delivery — no./total no. (%)
Both twins in cephalic presentation 798/1391 (57.4) 845/1393 (60.7)
First twin in cephalic presentation and second twin in noncephalic 542/1391 (39.0) 507/1393 (36.4)
presentation
First twin in noncephalic presentation and second twin in cephalic 51/1391 (3.7) 41/1393 (2.9)
or noncephalic presentation
Gestational age at delivery of first twin‡
Mean — wk 36.7±1.5 36.8±1.5
32 wk 0 days to 33 wk 6 days — no./total no. (%) 88/1392 (6.3) 66/1392 (4.7)
34 wk 0 days to 36 wk 6 days — no./total no. (%) 582/1392 (41.8) 581/1392 (41.7)
37 wk 0 days to 38 wk 6 days — no./total no. (%) 694/1392 (49.9) 696/1392 (50.0)
≥39 wk 0 days — no./total no. (%) 28/1392 (2.0) 49/1392 (3.5)
Time from randomization to delivery of first twin — days 12.4±12.0 13.3±12.3
Interval between deliveries — min 3.6±9.3 10.0±16.7
Use of antenatal glucocorticoids after randomization — no./ 179/1391 (12.9) 152/1393 (10.9)
total no. (%)
Use of anesthesia or analgesia — no./total no. (%)§ 1323/1390 (95.2) 996/1391 (71.6)
Regional 1226/1323 (92.7) 867/996 (87.0)
General 86/1323 (6.5) 53/996 (5.3)
Other 22/1323 (1.7) 109/996 (10.9)
Chorionicity at birth — no./total no. (%)¶
Dichorionic and diamnionic 1016/1390 (73.1) 1035/1391 (74.4)
Monochorionic and diamnionic 346/1390 (24.9) 324/1391 (23.3)
Monochorionic and monoamniotic 3/1390 (0.2) 1/1391 (0.1)
Unknown 25/1390 (1.8) 31/1391 (2.2)

* Plus–minus values are means ±SD. There were no significant between-group differences, except for gestational age at delivery (P = 0.01),
time from randomization to delivery of first twin (P = 0.04), and interval between deliveries (P<0.001). Data were missing for the following
characteristics: time from randomization to delivery of first twin (for two women in the planned-cesarean-delivery group and for one in the
planned-vaginal-delivery group) and interval between deliveries (for two women in the planned-cesarean-delivery group).
† Data include one singleton pregnancy that had been thought to be a twin pregnancy during ultrasonographic examination at randomization.
‡ Delivery between 37 weeks 5 days and 38 weeks 6 days of gestation occurred in 396 of 1392 women (28.4%) in the planned-cesarean-delivery
group and in 411 of 1392 (29.5%) in the planned-vaginal-delivery group.
§ Women could have received more than one type of anesthesia or analgesia.
¶ Chorionicity was confirmed by means of pathological examination at birth, with the following results: dichorionic and diamnionic (in 785 women
in the planned-cesarean-delivery group and in 768 in the planned-vaginal-delivery group), monochorionic and diamnionic (in 235 in the planned-
cesarean-delivery group and in 212 in the planned-vaginal-delivery group), and monochorionic and monoamniotic (in 3 in the planned-cesarean-
delivery group).

n engl j med 369;14 nejm.org october 3, 2013 1301

The New England Journal of Medicine
Downloaded from nejm.org on August 28, 2023. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

Table 3. Fetal or Neonatal Outcomes.

Planned Planned Odds Ratio

Outcome Cesarean Delivery Vaginal Delivery (95% CI) P Value
Fetal or neonatal death or serious neonatal morbidity
No. of fetuses or infants included in analysis 2783 2782
Composite primary outcome — no. (%)* 60 (2.2) 52 (1.9) 1.16 (0.77–1.74) 0.49
Gestational age at randomization — no./total no. (%)
32 wk 0 days to 33 wk 6 days† 32/948 (3.4) 26/956 (2.7) 1.25 (0.70–2.24)
34 wk 0 days to 36 wk 6 days 26/1358 (1.9) 19/1326 (1.4) 1.34 (0.71–2.54)
37 wk 0 days to 38 wk 6 days 2/477 (0.4) 7/500 (1.4) 0.30 (0.06–1.43)
Death — no. (%)‡ 24 (0.9) 17 (0.6)
Fetal death 13 (0.5) 9 (0.3)
Before the onset of labor 11 (0.4) 8 (0.3)
During delivery 0 1 (<0.1)
Unknown 2 (0.1) 0
Neonatal death 11 (0.4) 8 (0.3)
Serious neonatal morbidity — no. (%)‡ 36 (1.3) 35 (1.3)
Neonatal morbidity, excluding death of either twin
No. of infants included in analysis 2759 2765
Birth trauma — no. (%)
Long-bone fracture‡ 0 4 (0.1)
Other bone fracture 1 (<0.1) 1 (<0.1)
Facial-nerve injury at 72 hr of age or at discharge‡ 0 1 (<0.1)
Intracerebral hemorrhage‡ 3 (0.1) 1 (<0.1)
Apgar score <4 at 5 min — no. (%)‡ 2 (0.1) 7 (0.3)
Abnormal level of consciousness — no. (%)
Coma‡ 0 1 (<0.1)
Stupor or decreased response to pain‡ 2 (0.1) 0
Hyperalert, drowsy, or lethargic 9 (0.3) 7 (0.3)
≥2 Seizures within 72 hr after birth — no. (%)‡ 3 (0.1) 3 (0.1)
Assisted ventilation for ≥24 hr by means of endotracheal tube, 27 (1.0) 17 (0.6)
inserted within 72 hr after birth — no. (%)‡
Neonatal sepsis within 72 hours after birth — no. (%)‡ 1 (<0.1) 2 (0.1)
Necrotizing enterocolitis — no. (%)§ 1 (<0.1) 3 (0.1)
Cystic periventricular leukomalacia — no. (%)‡ 2 (0.1) 0

* The composite primary outcome was fetal or neonatal death or serious neonatal morbidity. The probabilities of interaction between treat­
ment group and baseline variables for the composite primary outcome were as follows: parity (0 vs. ≥1; P = 0.23); gestational age at random­
ization (32 weeks 0 days to 33 weeks 6 days, 34 weeks 0 days to 36 weeks 6 days, or 37 weeks 0 days to 38 weeks 6 days; P = 0.18); maternal
age (<30 years vs. ≥30 years; P = 0.63); presentation of second twin (cephalic vs. noncephalic; P = 0.51); chorionicity (dichorionic vs. mono­
chorionic; P = 0.15); and national perinatal mortality in the mother’s country of residence (<15 deaths per 1000 births, 15 to 20 deaths per
1000 births, or >20 deaths per 1000 births; P = 0.50). There were no infants with spinal cord injury, basal or depressed skull fracture, sub­
dural hematoma, meningitis, grade 3 or 4 intraventricular hemorrhage, or bronchopulmonary dysplasia in either group. CI denotes confi­
dence interval.
† Two infants (from one pregnancy) whose mother underwent randomization at 31 weeks of gestational age were included in the gestational-age
category of 32 to 33 weeks.
‡ This outcome was a component of the composite primary outcome.

1302 n engl j med 369;14 nejm.org october 3, 2013

The New England Journal of Medicine

Downloaded from nejm.org on August 28, 2023. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
 Planned Cesarean or Vaginal Delivery for Twins

Table 4. Maternal Outcomes.

Planned Planned Odds Ratio

Outcome Cesarean Delivery Vaginal Delivery (95% CI) P Value
Maternal death or serious maternal morbidity
No. of women included in the analysis 1392 1392
Death or serious maternal morbidity — no. (%) 102 (7.3) 118 (8.5) 0.86 (0.65–1.13) 0.29
Death — no. (%)* 1 (0.1) 1 (0.1)
Serious maternal morbidity, excluding death†
No. of women included in the analysis 1391 1391
Hemorrhage — no. (%)‡§ 84 (6.0) 108 (7.8)
Blood loss ≥1500 ml — no. (%) 26 (1.9) 32 (2.3)
Blood transfusion — no. (%) 66 (4.7) 75 (5.4)
Dilation and curettage of uterus after delivery 12 (0.9) 29 (2.1)
— no. (%)
Laparotomy — no. (%) 10 (0.7) 5 (0.4)
Genital tract injury — no. (%) 3 (0.2) 5 (0.4)
Perineal third- or fourth-degree tear involving anal sphincter 0 4 (0.3)
— no. (%)
Thromboembolism requiring anticoagulant therapy 5 (0.4) 2 (0.1)
— no. (%)
Infection, excluding wound infection — no. (%)¶ 25 (1.8) 18 (1.3)
Wound infection — no. (%)§ 27 (1.9) 18 (1.3)
Infection requiring prolongation of hospital stay 9 (0.6) 5 (0.4)
Infection requiring readmission to hospital 5 (0.4) 4 (0.3)
Infection requiring repeated treatment as an outpatient 14 (1.0) 9 (0.6)
Wound dehiscence or breakdown — no. (%) 20 (1.4) 12 (0.9)
Major serious or life-threatening medical illness 6 (0.4) 1 (0.1)
— no. (%)§
Acute respiratory distress syndrome 1 (0.1) 0
Disseminated intravascular coagulation 2 (0.1) 0
Amniotic-fluid embolism 1 (0.1) 0
Bowel obstruction or paralytic ileus requiring 3 (0.2) 0
nasogastric suction
Other serious or life-threatening illness‖ 1 (0.1) 1

* The causes of maternal deaths were cardiac failure during cesarean section due to a preexisting cardiomyopathy (in one woman in the
planned-cesarean-delivery group) and retroperitoneal hematoma after a massive postpartum hemorrhage requiring hysterectomy after
­vaginal delivery (in one woman in the planned-vaginal-delivery group).
† All measures of serious maternal morbidity listed were components of the composite maternal outcome. There were no cases of broad-
ligament hematoma confirmed by means of ultrasonography or other report or fistula involving the genital tract.
‡ Data were missing for one woman in the planned-cesarean-delivery group and for four women in the planned-vaginal-delivery group.
§ More than one subcategory may apply.
¶ Infection (excluding wound infection) was defined as a temperature of more than 38.5°C on two or more occasions at least 24 hours apart,
not including the first 24 hours after birth; pneumonia confirmed by means of radiography; or sepsis confirmed by means of blood culture.
‖ Other serious or life-threatening illnesses included generalized seizure (in one woman in the planned-cesarean-delivery group) and acute
fatty liver of pregnancy (in one woman in the planned-vaginal-delivery group).

n engl j med 369;14 nejm.org october 3, 2013 1303

The New England Journal of Medicine
Downloaded from nejm.org on August 28, 2023. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

SUBGROUP ANALYSES avoided selection bias, we ensured the presence of

There were no significant interactions for the an experienced obstetrician at delivery, and many
primary outcome between treatment group and of the twins in our study were born preterm.
parity (0 vs. ≥1; P = 0.23), gestational age at ran- We did not find any significant interactions
domization (32 weeks 0 days to 33 weeks 6 days, between treatment group and baseline variables,
34 weeks 0 days to 36 weeks 6 days, or 37 weeks suggesting no significant benefit of planned
0 days to 38 weeks 6 days; P = 0.18), maternal age cesarean delivery for any subgroup tested. How-
(<30 years vs. ≥30 years; P = 0.63), presentation ever, our study was not powered for these sub-
of the second twin (cephalic vs. noncephalic; group analyses. Further study may be warranted
P = 0.51), chorionicity (dichorionic vs. monocho- for the gestational-age subgroup of 37 to 38 weeks,
rionic; P = 0.15), or the national perinatal mortality particularly given the limited number of infants in
in the mother’s country of residence (<15 deaths this subgroup.
per 1000 births, 15 to 20 deaths per 1000 births, We did not find that planned cesarean delivery
or >20 deaths per 1000 births; P = 0.50). was associated with a higher or lower risk of
The second twin was more likely than the first maternal death or serious maternal morbidity
twin to have the primary outcome (odds ratio, than planned vaginal delivery. This finding may
1.90; 95% CI, 1.34 to 2.69, P<0.001). However, be explained in part by the high rate of cesarean
the interaction between treatment group and birth section (>40%) in the planned-vaginal-delivery
order was not significant (odds ratio for the first group, with most of these deliveries occurring
twin, 1.30; odds ratio for the second twin, 1.09; during labor.
P = 0.63). The strengths of our trial include the random-
ized design and use of central randomization, the
DISCUSSION large size of the study (106 participating centers
in 25 countries), and a high rate of follow-up.
In this large, randomized trial comparing deliv- Any possible unblinding of outcome assessors is
ery strategies for twins between 32 and 38 weeks unlikely to have introduced bias because the
of gestation, planned cesarean section did not criteria for the morbidity outcomes were clearly
reduce the risk of fetal or neonatal death or seri- defined. However, our findings are generaliz-
ous neonatal morbidity, as compared with planned able only to centers that can provide the obstetri-
vaginal delivery (with cesarean section if medi- cal management specified by the protocol, in-
cally indicated). We found a higher risk of an cluding the ability to perform an emergency
adverse perinatal outcome for the second twin cesarean section within 30 minutes if necessary.
than for the first twin, as others have found8-10; On the basis of the 95% confidence interval
however, planned cesarean section did not re- around the odds ratio for the primary outcome,
duce this risk. our results are consistent with no more than a
There has been controversy regarding the saf- 23% reduction and no more than a 74% increase
est method for the delivery of twins at or near in the odds of fetal or neonatal death or serious
term. A policy of planned cesarean section for the neonatal morbidity with planned cesarean deliv-
delivery of twins gained support after the publi- ery, as compared with planned vaginal delivery.
cation of the Term Breech Trial, which showed In conclusion, we found no benefits of planned
that planned cesarean delivery was associated cesarean section, as compared with planned vagi-
with a reduced risk of an adverse perinatal out- nal delivery, for the delivery of twins between 32
come in the case of a full-term pregnancy with and 38 weeks of gestation, if the first twin was in
the fetus in the breech presentation.17,22 Further the cephalic presentation.
support for planned cesarean section has come Supported by a grant (63164) from the Canadian Institutes of
from large cohort studies of twins showing a Health Research.
No potential conflict of interest relevant to this article was
reduced risk of an adverse perinatal outcome with reported.
elective cesarean section, as compared with vagi- Disclosure forms provided by the authors are available with
nal delivery or emergency cesarean section.7,8,10 the full text of this article at NEJM.org.
We thank all the participants in the Twin Birth Study and the
There are several possible reasons why our re- staff at the Centre for Mother, Infant, and Child Research for
sults differ from previous observational data: we their hard work and dedication.

1304 n engl j med 369;14 nejm.org october 3, 2013

The New England Journal of Medicine

Downloaded from nejm.org on August 28, 2023. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.
 Planned Cesarean or Vaginal Delivery for Twins

References

1. Wilcox LS, Kiely JL, Melvin CL, Martin twins at term: a retrospective cohort study 16. Persad VL, Baskett TF, O’Connell CM,
MC. Assisted reproductive technologies: of 8073 births. BJOG 2005;112:1139-44. Scott HM. Combined vaginal-cesarean de-
estimates of their contribution to multiple 9. Smith GC, Fleming KM, White IR. livery of twin pregnancies. Obstet Gynecol
births and newborn hospital days in the Birth order of twins and risk of perinatal 2001;98:1032-7.
United States. Fertil Steril 1996;65:361-6. death related to delivery in England, North- 17. Hannah ME, Hannah WJ, Hewson
2. Bergh T, Ericson A, Hillensjö T, ern Ireland, and Wales, 1994-2003: retro- SA, Hodnett ED, Saigal S, Willan AR.
Nygren KG, Wennerholm UB. Deliveries spective cohort study. BMJ 2007;334:576-8. Planned caesarean section versus planned
and children born after in-vitro fertilisa- 10. Armson BA, O’Connell C, Persad V, vaginal birth for breech presentation at
tion in Sweden 1982-95: a retrospective Joseph KS, Young DC, Baskett TF. Deter- term: a randomised multicentre trial.
cohort study. Lancet 1999;354:1579-85. minants of perinatal mortality and seri- Lancet 2000;356:1375-83.
3. Kiely JL. The epidemiology of perinatal ous neonatal morbidity in the second twin. 18. Su M, McLeod L, Ross S, et al. Factors
mortality in multiple births. Bull N Y Acad Obstet Gynecol 2006;108:556-64. associated with adverse perinatal outcome
Med 1990;66:618-37. 11. Lee HC, Gould JB, Boscardin WJ, El- in the Term Breech Trial. Am J Obstet Gy-
4. Ghai V, Vidyasagar D. Morbidity and Sayed YY, Blumenfeld YJ. Trends in cesar- necol 2003;189:740-5.
mortality factors in twins: an epidemio- ean delivery for twin births in the United 19. Pan W. Sample size and power calcu-
logic approach. Clin Perinatol 1988;15: States: 1995-2008. Obstet Gynecol 2011; lations with correlated binary data. Con-
123-40. 118:1095-101. trol Clin Trials 2001;22:211-27.
5. Smith GCS, Pell JP, Dobbie R. Birth 12. Arabin B, Kyvernitakis I, Liao A, 20. Klar N, Donner A. Current and future
order, gestational age, and risk of delivery Zugaib M. Trends in cesarean delivery for challenges in the design and analysis of
related perinatal death in twins: retrospec- twin births in the United States: 1995- cluster randomization trials. Stat Med
tive cohort study. BMJ 2002;325:1004-9. 2008. Obstet Gynecol 2012;119:657-8. 2001;20:3729-40.
6. Rabinovici J, Barkai G, Reichman B, 13. Cheung YB, Yip P, Karlberg J. Mortal- 21. Neonatal and perinatal mortality
Serr DM, Mashiach S. Randomized man- ity of twins and singletons by gestational country, regional and global estimates.
agement of the second nonvertex twin: age: a varying-coefficient approach. Am J Geneva: World Health Organization, 2004
vaginal delivery or cesarean section. Am J Epidemiol 2000;152:1107-16. (http://whqlibdoc.who.int/publications/
Obstet Gynecol 1987;156:52-6. 14. Imaizumi Y. Perinatal mortality in 2007/9789241596145_eng.pdf).
7. Hoffmann E, Oldenburg A, Rode L, twins and factors influencing mortality in 22. Hogle KL, Kilburn L, Hewson S, Gaf-
Tabor A, Rasmussen S, Skibsted L. Twin Japan, 1980-98. Paediatr Perinat Epidemiol ni A, Wall R, Hannah ME. Impact of the
births: cesarean section or vaginal delivery? 2001;15:298-305. international term breech trial on clinical
Acta Obstet Gynecol Scand 2012;91:463-9. 15. Puissant F, Leroy F. A reappraisal of practice and concerns: a survey of centre
8. Smith GC, Shah I, White IR, Pell JP, perinatal mortality factors in twins. Acta collaborators. J Obstet Gynaecol Can 2003;
Dobbie R. Mode of delivery and the risk Genet Med Gemellol (Roma) 1982;31: 25:14-6.
of delivery-related perinatal death among 213-9. Copyright © 2013 Massachusetts Medical Society.

2013 icmje recommendations

The International Committee of Medical Journal Editors (ICMJE) has revised
and renamed its Uniform Requirements. The new ICMJE Recommendations
for the Conduct, Reporting, Editing and Publication of Scholarly Work
in Medical Journals are available at www.icmje.org.

n engl j med 369;14 nejm.org october 3, 2013 1305

The New England Journal of Medicine
Downloaded from nejm.org on August 28, 2023. For personal use only. No other uses without permission.
Copyright © 2013 Massachusetts Medical Society. All rights reserved.