Background
Despite the demonstrated safety of a trial of labor for pregnancies with a vertex-presenting twin and clinical guidelines in support of this plan, the rate of planned cesarean delivery for twin pregnancies remains high. This high rate, as well as variation in cesarean rates for twin pregnancies across providers, may be influenced strongly by concern about delivery of the second twin, particularly when it is in a nonvertex presentation. There are limited data in the literature that has examined the impact of the position of the nonpresenting twin on successful vaginal delivery or maternal/neonatal morbidity.
Objective
We hypothesized that nonvertex presentation of the second twin would be associated with lower rates of successful vaginal birth for those patients attempting labor.
Study Design
This institutional review board–approved, retrospective cohort study of women who labored with twin pregnancies in a single urban hospital from 2007-2011. We included women with vertex-presenting first twins at >32 weeks gestation without a contraindication to labor and excluded those with uterine scar or lethal fetal anomaly. Vaginal delivery rates were evaluated according to vertex or nonvertex presentation of the second twin at admission and again at delivery. Maternal and neonatal morbidities were evaluated separately. Logistic regression was used to control for multiple confounders.
Results
Seven hundred sixteen patients met the inclusion criteria; 349 patients (49%) underwent a trial of labor. This included 73% (296/406) of eligible vertex/vertex twins and 17% (53/310) eligible vertex/nonvertex twins ( P < .01). When compared with laboring patients with vertex/vertex-presenting twins, those with vertex/nonvertex twins were younger (median age, 32 vs 33 years; P = .05), were more often multiparous (60% vs 43%; P = .02), and were less likely to have hypertension (13% vs 27%; P = .03). Eighty-five percent of patients with nonvertex second twins at admission delivered vaginally, compared with 70% of patients with vertex second twins ( P = .02). After we controlled for confounders, the difference was not statistically significant (adjusted odds ratio, 2.10; 95% confidence interval, 0.93–4.73). In the subset of patients with nonvertex second twins at delivery, those who initiated labor had an 89% vaginal delivery rate, compared with a 56% rate for those who changed from vertex to nonvertex presentation during labor (adjusted odds ratio, 19.90; 95% confidence interval, 3.86–102.78). Labor induction and increasing provider years in practice were also significant positive predictors of vaginal birth when the second twin was nonvertex at delivery. Maternal and neonatal morbidity was low and similar between groups, although 8% of women with nonvertex second twins experienced cervical lacerations, compared with 1% with vertex second twins ( P = .01).
Conclusion
Patients with nonvertex second twins had comparable, if not higher, rates of vaginal delivery than their vertex-presenting counterparts. The higher rate of vaginal delivery with stable nonvertex lie and the association with labor induction and the physician’s years in practice all suggest a role for provider selection and delivery planning. These findings and the observed 11% rate of intrapartum presentation change support vaginal delivery of the nonvertex second twin.
It is known that twin pregnancies carry an increased risk of neonatal and maternal morbidity compared with singleton gestations, but the specific impact of mode of delivery on outcome has remained a topic of debate. Although some observational studies have reported an increased risk of morbidity for vaginally delivered second twins, more recent observational studies, meta-analyses, and decision analyses support the safety of vaginal delivery. In addition, a recent large randomized trial, the Twin Birth Study, demonstrated that a trial of labor was associated with improved maternal and infant outcomes when compared with planned cesarean delivery. Based on these findings, consensus guidelines from the American Congress of Obstetricians and Gynecologists and the Society for Maternal-Fetal Medicine suggest counseling patients with cephalic-presenting twins for an attempt at vaginal birth.
Despite the demonstrated safety of a trial of labor for pregnancies with a vertex-presenting twin, the rate of planned cesarean delivery for twin pregnancies remains 44% in a recent prospective cohort study. This high rate, as well as variation in cesarean rates for twin pregnancies across providers, may be influenced strongly by concern about delivery of the second twin, particularly when in a nonvertex presentation.
There are limited data in the literature that have examined the impact of the position of the nonpresenting twin on successful vaginal delivery or on maternal/neonatal morbidity. We undertook this study to examine this question and hypothesized that nonvertex presentation of the second twin would be associated with lower rates of successful vaginal birth for those patients who attempt labor.
Methods and Materials
Approval was obtained from the Partners Human Research Committee. Using hospital birth records, we identified all women with twin pregnancies of at least 32 and 0/7 weeks gestation who delivered at Brigham and Women’s Hospital between January 2007 and December 2011. Pregnancies at <32 weeks gestation were excluded for analysis based on the absence of safety data for breech extraction and the suggestion of harm for nonvertex second twins who were delivered vaginally at an extremely premature gestational age.
Brigham and Women’s Hospital is an urban, academic, tertiary care center with an obstetric staff of 8 provider groups that are composed of approximately 90 different clinicians. Labor management and choice of delivery method in this retrospective cohort were at the discretion of the delivering provider because the hospital does not have specific protocols regarding the intrapartum management of twins.
Because we wished to determine the proportion of eligible women who underwent a trial of labor, we excluded women with contraindications to labor, such as those with placenta previa, a nonvertex presenting twin, previous transmural uterine surgery, abdominal cerclage, monoamniotic gestation, category 3 fetal heart rate tracing at admission, or nonlethal fetal anomalies precluding labor or vaginal birth (ie, neural tube defect). We also excluded patients with a previous cesarean delivery, an intrauterine fetal death, or lethal fetal anomaly. Patients were then categorized as vertex/vertex (VV) or vertex/nonvertex (VNV), according to the documented admission presentation of the second twin. A nonvertex second twin was defined as either breech or transverse (ie, noncephalic) presentation.
Definition of variables
Obstetric records were reviewed to obtain data regarding maternal, obstetric, fetal, and provider factors. Maternal body mass index (BMI) was recorded at the time of delivery, because the prepregnancy BMI was unavailable for a large number of patients. Hypertension was defined when a provider diagnosed chronic or gestational hypertension (including preeclampsia). Maternal diabetes mellitus was considered present if it required medication (either gestational or pregestational). Other maternal medical conditions included chronic cardiac, neurologic, or hematologic disease. A patient with hypertension, diabetes mellitus, or other maternal medical morbidities as defined earlier was considered to have a medical condition for this analysis.
Chorionicity, estimated fetal weights, weight percentiles, and presence of fetal anomalies were transcribed directly from patients’ ultrasound records. To restrict clinical predictors to those available to providers who managed the deliveries, information was transcribed directly from sonographic reports without normalization of standard growth curves. Registered Diagnostic Medical Sonographers performed ultrasound scans that were reviewed by board-certified radiologists or maternal-fetal medicine specialists. Sonographic assessment of fetal weight was transcribed directly from the last antepartum ultrasound report; intrauterine growth restriction was defined as an estimated fetal weight below the tenth percentile. Chorionicity was defined by standard sonographic definitions. We did not standardize growth curves for the purpose of this study because we wished to use the exact weight percentiles that were reported to the delivering clinicians. Weight discordance was calculated by dividing the weight difference by the larger twin’s estimated fetal weight; a discrepancy of >20% used to define weight discordance . Given that no specific guidelines on weight discordance and the safety of vaginal birth exist either in the literature or in our hospital’s protocols, it was not an exclusion criterion for this study. Ultrasound scans to confirm presentation of the presenting fetus were performed at the time of admission, as is mandated by department policy. Although not required by policy, ultrasound scanning that is performed by the delivering obstetrician to confirm presentation at the time of delivery is routinely, if not universally, used for twin gestations.
Obstetric data included labor induction status and the patient’s cervical dilation at the time of admission. Labor was considered induced if any drug or mechanical method (including amniotomy) was used to initiate cervical dilation or contractions. For each physician provider, we determined the level of training (Maternal Fetal Medicine specialist or generalist) and year of residency graduation. Certified nurse midwives were permitted to deliver twin patients under the supervision of a physician and were categorized as generalist providers for this study. If an attending physician directly assisted the midwife, then the physician’s characteristics were considered for provider analyses. The physician’s number of years in practice was calculated from the time of residency graduation until the date of delivery, with midwife deliveries excluded from this calculation and the analyses using this variable.
Definition of outcomes
The primary outcome of interest was vaginal delivery of both twins among women who attempted a vaginal delivery. For the purpose of this analysis, patients who encountered a vaginal delivery of the presenting twin and a cesarean delivery of the second twin, also referred to combined delivery, were classified as having had a cesarean delivery.
Additional outcomes included maternal and neonatal morbidity according to fetal presentation at admission. Maternal morbidities included combined vaginal and cesarean delivery, intrapartum fever (defined as a temperature ≥38°Celsius) or chorioamnionitis (fever plus use of intrapartum antibiotics), need for general anesthesia, postpartum hemorrhage (defined as an estimated blood loss of over 1 liter or need for red blood cell transfusion), postpartum infection (endometritis with use of postpartum antibiotics, wound or other infection that required antibiotics, or need for wound reopening or abscess drainage), and readmission within 6 weeks after delivery.
Rates of major procedures and lacerations were recorded. Major procedures included hysterectomy, laparotomy, uterine artery embolization, placement of an intraabdominal or pelvic drain; major lacerations included cervical or third- or fourth-degree perineal lacerations. Readmission within the 6-week postpartum period was included only if it was related to a potential delivery complication (such as a venous thromboembolism, infection, or hemorrhage) and not related to blood pressure, glycemic control, or a new medical condition.
Neonatal outcomes were evaluated for the pregnancy overall and separately for presenting and nonpresenting twins. Adverse outcomes included 5-minute APGAR scores < 7, neonatal intensive care unit admission for births at ≥36 weeks gestation, birth trauma, need for blood transfusion, infection that required antibiotics, intraventricular hemorrhage, and death. These diagnoses were transcribed directly from the neonates’ medical records with diagnostic confirmation of birth trauma and intraventricular hemorrhage based on the evaluation of the pediatric team. Composite morbidity was defined for each twin, with a newborn infant counted once if it had any of the aforementioned morbidities, with the exception of a low APGAR score or neonatal intensive care unit admission, neither of which represent direct physical harm. Composite morbidity for a single pregnancy was defined as either twin having a positive individual composite morbidity.
To better understand predictors of vaginal birth for nonvertex second twins, we then examined outcomes among those women with nonvertex second twin at the time of delivery. Patients were considered as having a nonvertex second twin at delivery if the presentation of the second twin was anything other than cephalic after delivery of the first twin. If the second twin was turned actively to a new presentation before birth (by either external or internal version), the presentation before the version was considered. Women with a nonvertex second twin at both admission and delivery were categorized as having stable nonvertex second twins. Those who initiated labor with vertex second twins, but ended with a second twin in nonvertex presentation, were categorized as unstable-lie nonvertex second twins.
Statistical analysis
The associations between admission presentation and maternal, fetal, obstetric, or provider characteristics were evaluated with the chi square test or Fishers exact test for categoric variables. The Wilcoxon rank sum test was used for continuous variables, because all of the continuous variables fell into a nonnormal distribution. Statistical significance was defined with a probability value of <.05. Logistic regression models examined the associations while controlling for potentially confounding variables. Potential confounders for inclusion in the model were identified by the determination of whether they changed the odds ratio for the outcome by >10%. A backwards stepwise procedure was used to arrive at the final model.
Because our outcome occurred at a high rate, large calculated odds ratios (>2.5) were converted to estimated relative risks according to the method published by Zhang and Yu. Study data were collected and managed with the use of REDCap electronic data capture tools hosted at Harvard University. All analyses were performed with Statistical Analysis Software (version 9.3; SAS Institute, Inc, Cary, NC). With a base vaginal delivery rate of 72% and an alpha level of 0.05 (2-sided), we had 88% power to detect a 30% lower vaginal delivery rate with nonvertex twins.
Results
Figure 1 shows the identification and categorization of patients for the current study. Of 1250 patients who delivered twins at ≥32 weeks gestation, 716 patients (58%) were eligible for a trial of labor and met inclusion criteria for the study with complete records for review. Of this group, 406 patients (57%) had twins in VV presentation; the remaining 310 patients (43%) had VNV presentations at admission. The VV and VNV groups did not differ significantly in terms of maternal demographics, medical history, gestational age at delivery, presence of fetal anomalies, or growth restriction. There was a trend towards more assisted conceptions (64% vs 57%; P = .08) and more estimated weight discordance with a larger first twin (8% vs 5%; P = .09) in the VNV group compared with the VV group. Notably, only 17% of all patients with nonvertex second twins (53/310) attempted labor compared with 73% of patients with vertex second twins (296/406; P < .0001). The 349 women (49%) who attempted labor formed the cohort for the current study, of whom 296 women (85%) had VV twins; the other 53 women (15%) twins in VNV presentation.
Eighty-four different providers delivered this group of laboring patients. Twenty-two of the providers (26%) were Maternal Fetal Medicine specialists, who delivered 121 (35%) of the patients. Eleven deliveries (3%) were performed by 8 different certified nurse midwives; 5 of the midwives were assisted directly by an attending physician. The remaining 217 patients (62%) were delivered by 54 different generalist obstetricians. Table 1 summarizes the baseline patient and delivery characteristics for women who elected a trial of labor according to admission presentation of the second twin. When compared with laboring patients with VV presenting twins, those with VNV presenting twins were younger (median age, 32 vs 33 years; P = .05), more often multiparous (60% vs 43%; P = .02), and less likely to have hypertension (13% vs 27%; P = .03).
| Characteristic a | Vertex/vertex (n = 296; 85%) | Vertex/nonvertex (n = 53; 15%) | P value b |
|---|---|---|---|
| Median maternal age, y (range) c | 33 (18–48) | 32 (19–43) | .05 |
| Multiparous, n (%) | 127 (43) | 32 (60) | .02 |
| White race, n (%) | 204 (69) | 32 (60) | .22 |
| Median body mass index at delivery, kg/m 2 (range) c | 30 (20–55) | 30 (19–49) | .25 |
| Any medical condition, n (%) | 92 (31) | 11 (21) | .13 |
| Hypertension, n (%) | 80 (27) | 7 (13) | .03 |
| Spontaneous conception, n (%) | 144 (49) | 32 (60) | .12 |
| Monochorionic, n (%) | 56 (19) | 9 (17) | .74 |
| Median gestational age, wk (range) c | 36.6 (32–40) | 36.9 (32–40) | .48 |
| Induced labor, n (%) | 192 (65) | 30 (57) | .25 |
| Admission dilation ≥3cm, n (%) | 150 (51) | 33 (62) | .14 |
| ≥20% Discordance, n (%) | 24 (8) | 6 (12) | .44 |
| Twin A→twin B | 7 (2) | 4 (8) | .07 |
| Twin B→twin A | 17 (6) | 2 (4) | .75 |
| Intrauterine growth restriction of either twin, n (%) | 37 (13) | 5 (9) | .65 |
| Nonlethal anomaly, n (%) | 16 (5) | 0 | .14 |
| Maternal-fetal medicine provider, n (%) | 101 (34) | 20 (38) | .61 |
| Median practice time of provider, n (range) c , d | 10 (0-35) | 11 (0-28) | .45 |
a Admission dilation missing for 4 subjects; estimated fetal weight for discordance missing for 7 subjects; intrauterine growth restriction missing for 6 subjects
b Calculated by chi square or Fisher exact tests, unless otherwise indicated
c Probability value determined by Wilcoxon-Rank Sum tests
d Provider refers to the physician performing the delivery, with years calculated from the time of residency graduation to the time of the delivery; 6 deliveries that were performed exclusively by midwives were excluded from this analysis, all of which were in the vertex/vertex group.
Additionally, patients with VNV twins more often had a second twin that was estimated to be >20% smaller than the presenter, although this did not reach statistical significance (8% discordance vs 2% in the VV group; P = .07).
Overall, 72% of 251 patients achieved a vaginal birth; 28% (n = 98) delivered by cesarean delivery. In univariate analyses ( Supplementary Table 1 ), vaginal delivery was significantly related to parity and to cervical dilation at admission. The 84% vaginal delivery rate for multiparous women was significantly higher than the 62% rate for primiparous patients ( P < .0001). Furthermore, those women who achieved a vaginal birth were significantly more likely to start labor with a cervical dilation of ≥3 cm (41% vs 31%; P < .01), although induction of labor was not associated with mode of delivery. We observed a trend towards more cesarean deliveries for Maternal Fetal Medicine providers, with a 66% vaginal delivery rate for this group, compared with a 75% vaginal birth rate for generalist and midwife providers ( P = .08).
Table 2 presents primary and Table 3 presents secondary outcomes according to admission presentation. Overall, 85% of patients with VNV twins at admission experienced vaginal birth, which was a significantly higher rate than the 70% rate for the VV twins ( P = .02). In a logistic regression analysis that controlled for parity, the odds ratio for vaginal birth in the VNV group was 2.10 (95% confidence interval, 0.93–4.73), but the difference was no longer statistically significant.
| Vaginal delivery | Admission presentation, n/N (%) | P value a | Adjusted odds ratio (95% confidence interval) | |
|---|---|---|---|---|
| Vertex/vertex | Vertex/nonvertex | |||
| All subjects | 207/296 (70) | 45/53 (85) | .02 | 2.10 (0.93–4.73) |
| Nonvertex second twin at delivery b | 19/34 (56) | 33/37 (89) | < .001 | 19.90 (3.86–102.78) |
a Calculated by chi square or Fisher exact tests
b Vaginal delivery for those with nonvertex twins at delivery (n = 71; vertex/vertex group, 34, vertex/nonvertex group, 37; relative risk, 1.72.
| Outcomes | Admission presentation | P value a | |
|---|---|---|---|
| Vertex/vertex (n = 296; 85%) | Vertex/nonvertex (n = 53; 15%) | ||
| Delivery, n (%) b | |||
| Combined delivery | 30 (10) | 4 (8) | .79 |
| Operative vaginal delivery | 33 (11) | 7 (13) | .66 |
| Breech extraction | 15 (5) | 29 (55) | < .001 |
| Maternal | |||
| Intrapartum fever, n (%) | 25 (8) | 6 (11) | .44 |
| General anesthesia, n (%) | 8 (3) | 1 (2) | 1.0 |
| Postpartum hemorrhage, n (%) | 46 (16) | 9 (17) | .84 |
| Estimated blood loss (cc) c | 575 (200–6000) | 500 (200–2500) | .67 |
| Transfusion, n (%) | 25 (8) | 6 (11) | .44 |
| Any postpartum infection, n (%) | 31 (11) | 5 (9) | 1.0 |
| Endometritis | 29 (10) | 5 (9) | 1.0 |
| Wound/skin infection | 3 (1) | 0 | 1.0 |
| Other infection | 1 (0.3) | 0 | 1.0 |
| Major laceration, n (%) | 8 (3) | 4 (8) | .09 |
| Third- or fourth-degree perineal | 5 (2) | 0 | 1.0 |
| Cervical | 3 (1) | 4 (8) | .01 |
| Major postpartum procedure, n (%) d | 8 (3) | 0 | 1.0 |
| Postpartum readmission, n (%) | 6 (2) | 1 (2) | 1.0 |
a Calculated by chi square or Fisher exact tests, unless otherwise indicated
b Vaginal delivery for those with nonvertex twins at delivery (n = 71; vertex/vertex group, 34, vertex/nonvertex group, 37; relative risk, 1.72
c Probability value determined by Wilcoxon-Rank Sum tests
d Major procedures include laparotomy, hysterectomy, intraabdominal or pelvic drain placement, or uterine artery embolization.
Indications for cesarean delivery included nonreassuring fetal heart rate (37%), failure to progress in the first stage of labor (20%), failure to progress in the second stage of labor (27%), breech presentation of the nonpresenting twin (10%), and other indications (6%). The 10 patients who underwent cesarean delivery for an indication of breech were in the VV group at admission. The risk of a combined route of delivery was similar in the VNV and VV groups (8% and 10%, respectively; P = .79). Operative vaginal delivery rates were similar in the 2 groups (11% vs 13%; P = .66); the VNV group underwent an expected higher rate of breech extraction (55% vs 5%; P < .0001).
The only significant difference in maternal morbidity was a higher rate of cervical laceration in the VNV group (8% vs 1% for the VV group; P = .01). To ensure that this was not a reflection of a higher vaginal delivery rates for this group, we restricted this analysis to just those women who achieved a successful vaginal birth, and the association remained (9% of VNV twins with vaginal delivery vs 1% for VV vaginal births; P = .01).
Rates of neonatal morbidity and mortality ( Table 4 ) were low and similar between the groups, both for specific outcomes and for the composite morbidity measures for the pregnancy overall. Individual outcomes for presenting and nonpresenting twins were also low and similar between VV and VNV groups ( Supplemental Table 2 ). The median intertwin delivery time interval was significantly lower in the VNV group (5 vs 12 minutes; P < .0001). One second twin in the VV group experienced a femur fracture after starting from vertex presentation but delivered as a vaginal breech after an intrapartum change in presentation. The maternal labor course was uncomplicated, but this twin had overwhelming Escherichia coli bacteremia that was attributed to urosepsis, and died on day of life 5 as a result of this disease.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
