Objective
The optimal gestational duration for twin gestations is unknown. Epidemiologic studies show that the lowest perinatal mortality rate for twins is at 37-38 weeks, but these studies lack information on pregnancy complications and neonatal morbidities. This study evaluates pregnancy characteristics and perinatal outcomes of twins in order to assess the optimal gestational age for delivery.
Study Design
This is a retrospective study of twins delivered at ≥36 weeks at our institution from 1991-2009. The composite rate of perinatal morbidity and mortality (including perinatal death, respiratory distress, suspected sepsis, and need for neonatal intensive care) was determined for weekly intervals from 36-39 + weeks.
Results
There were 377 twin gestations included. Of those 83% were dichorionic. Fifty-three percent had spontaneous labor and 48% were delivered by cesarean section. Perinatal outcomes improved as gestational age advanced to 38 weeks.
Conclusion
Perinatal morbidity and mortality rates suggest that the optimal time for delivery of twins is at 38 weeks or greater.
The optimal duration of pregnancy for twin gestations is unknown and may be shorter than that of singletons. Several large epidemiologic studies have shown that the lowest perinatal mortality rate for twins occurs at an earlier gestational age (GA) than singletons. A study of data from the US National Center for Health Statistics, including 11.1 million singleton and 297,622 twin births, showed that the lowest perinatal mortality for singletons was at 39 to 41 weeks, but for twins was at 37 to 38 weeks. Moreover, as of 39 weeks, the prospective risk of fetal death in an ongoing twin pregnancy exceeded the risk of neonatal death, suggesting that delivery by 39 weeks may improve perinatal outcome. Another study of 88,936 infants of multiple gestations (96% twins) from Japan found that the incidence of stillbirth and early neonatal death gradually declined until 37-38 weeks and then increased. These same outcomes for singletons declined until 39 weeks before increasing. A review of the Swedish Medical Birth Registry including 32,942 twins showed that perinatal mortality rates were lowest at 37-38 weeks. Other smaller epidemiologic studies have also shown that the lowest perinatal mortality rates for twins are at 37-38 weeks.
These epidemiologic studies indicate that the lowest perinatal mortality rates for twin pregnancies occur at 37-38 weeks. However, because of their epidemiologic nature, these studies are limited by lack of information about pregnancy dating, chorionicity, and pregnancy complications. In addition, they lack information about the important outcome of neonatal morbidities. Previous studies have addressed neonatal morbidities, but have revealed conflicting results. Some studies have suggested that twins mature faster than singletons, and, therefore, may be better equipped for earlier delivery. Other studies have shown increased neonatal morbidity for twins delivered before 38 weeks when compared with later delivery. Because of this conflicting information about neonatal morbidities, the limitations of the epidemiologic studies at evaluating perinatal mortality, and the lack of major randomized controlled trials, there is no consensus on the optimal gestational length for twins. This study was designed to examine the perinatal morbidity and mortality of twins delivered at or near term at our institution to determine the optimal gestational duration for twin pregnancies.
Materials and Methods
This is a retrospective cohort study that was approved by our Institutional Review Board. Our electronic obstetric record was queried to identify twin gestations that delivered at the University of Alabama, Birmingham, from 1991-2009. Only those twins delivered at 36 completed weeks or beyond were included. Exclusion criteria were major fetal anomalies, aneuploidy, and death of 1 or both twins before 36 weeks. Deliveries were categorized according to number of completed weeks of gestation. GA was determined by obstetric providers with the use of standard criteria that took into consideration the clinical history and the results of the earliest ultrasound (US) examination. If the findings on US examination were consistent with a GA based on a certain date of the last menstrual period (LMP), GA was determined according to the date of the LMP; if the date of the LMP was uncertain or the findings on US examination were inconsistent with a GA based on the date of the LMP, GA was determined according to the results of the US examination. Specifically, menstrual dating was used if the US measurements were ±7 days from the LMP before 22 weeks; if the US measurements were ±10 days from the LMP at 22-24 6/7 weeks; and if the US measurements were ±14 days from the LMP above 25 weeks.
Maternal and infant chart review was used to obtain demographic information, pregnancy complications, and pregnancy outcomes. We collected information about the type of conception (spontaneous vs assisted reproduction), maternal medical conditions, smoking status, and obesity rates (body mass index [BMI] >30 kg/m 2 at the initial antepartum visit). We also collected information about pregnancy and delivery complications including chorioamnionitis, hypertensive disorders, prior cesarean delivery, fetal growth restriction (FGR) (estimated fetal weight <10th% for GA based on data by Brenner et al ), oligohydramnios (greatest vertical pocket of ≤2.0 cm), polyhydramnios (greatest vertical pocket of >8 cm), and meconium-stained amniotic fluid. Chorioamnionitis was determined clinically based on maternal temperature of >100.4°F, maternal or fetal tachycardia, and uterine tenderness. Ampicillin and gentamicin were administered as per our standard treatment protocol for chorioamnionitis. Gestational hypertension, preeclampsia, and chronic hypertension were defined as per guidelines from the American College of Obstetricians and Gynecologists. Chorionicity was determined by US criteria and/or pathologic review of fetal membranes (when available on chart review). If the mother was given antenatal corticosteroids during the pregnancy, the name of the steroid and number of doses were recorded. The type of labor was determined for each patient (induction vs spontaneous vs none for those with planned cesarean delivery) and the indication for induction of labor was recorded. Mode of delivery and indication for cesarean delivery were determined. A notation was made regarding the GA at initiation of antepartum fetal surveillance and the type of testing received during the pregnancy.
Perinatal morbidity and mortality rates for twins delivered at each week of GA were collected. Thirty-nine to 41 week gestations were grouped into 1 group because of low numbers of deliveries beyond 39 weeks (n = 6). The primary outcome was the composite rate of perinatal morbidity and mortality (including perinatal death, any respiratory distress, suspected sepsis, and neonatal intensive care unit [NICU] admission) at weekly intervals from 36 to 39 + weeks. Perinatal death rates included stillbirths and neonatal deaths up to 28 days. Infants having “any respiratory distress” were those admitted to the NICU with a diagnosis of respiratory distress given by the admitting physician. A notation was made of the type of respiratory support administered, if any was required. The diagnosis of respiratory distress syndrome (RDS) required signs of respiratory distress, consistent radiologic features, and oxygen therapy with a fraction of inspired oxygen (FIO 2 ) of 0.40 or greater for at least 24 hours. Infants had “suspected sepsis” if they had clinical findings suggesting infection and received a sepsis work-up including blood and/or cerebrospinal fluid cultures and were given broad-spectrum antibiotics. Twin perinatal outcomes were compared with a matched group of singletons from our institution in order to determine whether twins appear to mature earlier than singletons, ie, experience lower morbidity than singletons at a given GA. Twins were matched at a 1:1 ratio of randomly selected singletons by year of delivery, GA week at delivery, race, and gender of fetus.
Statistical analysis for this study was performed using SAS statistical software (version 9.1; SAS Institute Inc., Cary, NC). Rates of the composite primary outcome and its components were compared using χ 2 , Fisher exact tests, and tests of trend. These same methods were used to compare outcomes of twins with singletons. Regression analysis was performed using general estimating equation framework to control for twin cluster correlations. Odds ratios (OR) were determined and adjusted for factors that were most likely to affect perinatal outcome which included race, chorionicity, FGR, diabetes, chorioamnionitis, steroid use, induction of labor, and delivery mode. To further evaluate the role of chorionicity on outcomes, the Breslow-Day test for homogeneity was performed after stratifying by twin chorionicity. Alpha was set at .05.
Results
Of the 435 twin gestations initially identified from our electronic obstetric record, 58 were excluded (31 for unknown GA or GA <36 weeks, 18 for anomalies, 2 for fetal demise of 1 or both twins <36 weeks, 2 for delivery outside of our institution, 1 for higher order multiple gestation, 3 for multiple exclusions, and 1 for unknown reason). Three hundred seventy-seven twin gestations (754 infants) remained that met inclusion and exclusion criteria. Table 1 shows patient demographics. The twin gestations were dichorionic in 83% of cases and monochorionic in 17%, with 0.8% of all gestations being monoamniotic. The mean GA at delivery was 37.5 weeks. There were 5 deliveries at 40 weeks and 1 delivery at 41 weeks.
| Demographics | N = 377 (%) |
|---|---|
| Maternal age, y | 25.9 a ± 6.3, 14–49 b |
| Race | |
| African American | 223 (59.2) |
| White | 120 (31.8) |
| Other | 34 (9.0) |
| Insurance | |
| Public | 280 (74.3) |
| Private | 64 (17.0) |
| None | 25 (6.6) |
| Unknown | 8 (2.1) |
| Type of conception | |
| Spontaneous | 353 (93.6) |
| Assisted | 23 (6.1) |
| Unknown | 1 (0.3) |
| Maternal medical and obstetric complications | |
| Diabetes | |
| Gestational | 21 (5.6) |
| Preexisting | 6 (1.6) |
| Chronic hypertension | 25 (6.6) |
| Smoking | 74 (19.7) |
| Obesity (BMI >30 kg/m 2 ) | 131 (34.7) |
| Gestational hypertension | 57 (15.1) |
| Preeclampsia | |
| Mild | 52 (13.8) |
| Severe | 7 (1.9) |
| Prior cesarean delivery | 60 (15.9) |
| Fetal growth restriction | 43 (11.4) |
| Oligohydramnios | 31 (8.2) |
| Polyhydramnios | 3 (0.8) |
| Premature ruptured membranes | 17 (4.5) |
| Neonatal characteristics | |
| Mean birthweight, g | 2594 ± 416 |
| Apgar score <7 at 5 min | 21 (2.8) |
| Small for gestational age | 193 (25.6) |
| Neonatal intensive care admission | 129 (17.1) |
| Suspected sepsis | 83 (11.0) |
| Respiratory distress | 76 (10.1) |
Fifty-six percent of women (210 of 377 twin gestations) had an uncomplicated pregnancy, not affected by a maternal medical condition or obstetric complication. For the 44% of women with a complication, their coexistent medical conditions and obstetric complications are shown in Table 1 . There were no elective deliveries—all patients had spontaneous labor or a medically indicated delivery. Specifically, 52.5% of the twin pregnancies had spontaneous labor, 24.2% had induction of labor, and 23.3% had no labor. Of the 91 patients who underwent labor induction, the 3 most common indications were preeclampsia or gestational hypertension (42 patients, 46.2%), suspected FGR (25 patients, 27.4%), and other maternal medical complications (21 patients, 23.1%). Delivery was vaginal (operative or spontaneous) for 52% and by cesarean for 48% of infants. The 2 most common indications for cesarean delivery were malpresentation (58%) or repeat cesarean delivery (27%). Intrapartum complications were rare—the most frequent were meconium-stained amniotic fluid (7%) and chorioamnionitis (4%). Ten percent of patients received betamethasone at some point during the pregnancy and the average number of doses was 2 injections. Ninety-five percent of patients received antepartum fetal monitoring starting at a mean of 33 weeks and contraction stress tests and nonstress tests were the most commonly used methods. There were no cases of twin-twin transfusion syndrome that met the inclusion criteria; all were excluded for meeting 1 or more of the exclusion criteria.
The neonatal characteristics of the 754 infants are shown in Table 1 . The 3 most common reasons for NICU admission were respiratory distress or transient tachypnea (91 infants), suspected sepsis (83 infants), and small for gestational age (SGA) (42 infants). Some of the infants had more than 1 indication for NICU admission. There were 6 perinatal deaths (0.8% of all infants) with causes including unknown (1 loss at 38 weeks), a cord accident (at 37 weeks), loss of a single fetus in a monochorionic pair (at 36 weeks), and loss of both fetuses from uterine rupture after a motor vehicle collision (at 36 weeks). Finally, a SGA male infant of a dichorionic twin pair, born at 40 weeks 0 days, died on day of life 5 because of persistent pulmonary hypertension. His amniotic fluid was meconium-stained.
Univariate analysis of the maternal diseases and obstetric conditions is shown in Table 2 and of the perinatal outcomes is shown in Table 3 . The rates of the composite perinatal outcome, NICU admission, suspected sepsis, any respiratory distress, and hyperbilirubinemia decreased as GA advanced. Logistic regression analysis showed that our primary outcome (the composite rate of perinatal death, suspected sepsis, respiratory distress, and NICU admission) was lowest at 38 weeks’ gestation. At 38 weeks, perinatal outcomes were significantly better than at 36 or 37 weeks, but were not statistically significantly different from 39 + weeks. Table 4 shows the corresponding ORs. When stratified by chorionicity, the Breslow-Day test for homogeneity showed no significant differences in results for all of the outcomes between monochorionic and dichorionic twins ( P values ranging from .19 to .70). When outcomes were evaluated for only the uncomplicated pregnancies (210 women with no medical or obstetric complications), there was an increased risk for the composite outcome and NICU admission at 36 weeks compared with 38 weeks, OR, 3.1; 95% confidence interval, 1.2–7.8 and OR, 3.5; 95% confidence interval, 1.3–9.3 for these outcomes, respectively. Results for 37 and 39 + weeks were not statistically different from 38 weeks.
| Condition | 36 wks, % | 37 wks, % | 38 wks, % | 39 + wks, % | P value |
|---|---|---|---|---|---|
| (n = 130) | (n = 118) | (n = 84) | (n = 45) | trend | |
| Diabetes | 11.5 | 5.9 | 6.0 | 0 | < .01 |
| Chronic HTN | 5.4 | 11.0 | 4.8 | 2.2 | .43 |
| Gestational HTN | 13.1 | 17.8 | 15.5 | 13.3 | .85 |
| Preeclampsia | 19.2 | 19.5 | 10.7 | 6.7 | .02 |
| Steroids received | 16.9 | 10.2 | 6.0 | 0 | < .01 |
| Spontaneous labor | 59.2 | 53.4 | 47.6 | 40.0 | .02 |
| Cesarean section | 43.9 | 50.9 | 50.0 | 46.7 | .54 |
| Growth restriction | 15.4 | 11.9 | 8.3 | 4.4 | .03 |
| Oligohydramnios | 6.2 | 5.9 | 13.1 | 11.1 | .08 |
| Polyhydramnios | 1.5 | 0.9 | 0 | 0 | .18 |
| Meconium | 4.6 | 4.3 | 9.5 | 13.3 | .02 |
| Perinatal outcome | 36 wks, % | 37 wks, % | 38 wks, % | 39 + wks, % | P value |
|---|---|---|---|---|---|
| (n = 260) | (n = 236) | (n = 168) | (n = 90) | ||
| Composite a | 30.0 | 15.7 | 7.1 | 7.8 | < .01 |
| NICU admission | 28.9 | 15.3 | 6.6 | 7.8 | < .01 |
| Suspected sepsis | 18.5 | 9.3 | 5.4 | 4.4 | < .01 |
| Any respiratory distress | 16.2 | 9.8 | 3.6 | 5.6 | < .01 |
| Respiratory support | 5.4 | 4.2 | 2.4 | 2.2 | .41 |
| Transient tachypnea | 1.9 | 2.1 | 3.0 | 3.3 | .76 |
| Hyperbilirubinemia | 5.8 | 1.3 | 0.6 | 0.0 | < .01 |
| SGA | 31.9 | 19.0 | 25.5 | 32.6 | .71 |
| Perinatal death | 1.2 | 0.4 | 0.6 | 1.1 | .79 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
