Introduction

As demonstrated by National Vital Statistics, the overwhelming majority of triplet gestations in the United States are delivered by planned cesarean delivery. However, evidence supporting an optimal route of delivery for triplet gestations is limited. The published literature to date contains conflicting data, with some studies reporting improved perinatal outcomes for planned cesarean delivery, while others suggest equivalent to improved outcomes with vaginal delivery. Overall, this body of literature on mode of delivery for triplet gestations has significant limitations, including: overrepresentation by small, single-center uncontrolled case series; use of noncontemporaneous historical controls; dissimilarity between vaginal and cesarean delivery groups on factors relevant to perinatal outcome (eg, gestational age); and comparison of outcomes of the actual, rather than the intended or planned, route of delivery. The largest study of triplet mode of delivery, a population-level analysis of deliveries >24 weeks gestation by Vintzileos et al, demonstrated an increased risk of gestational age–adjusted neonatal and infant mortality with vaginal compared with cesarean delivery. However, the use of administrative data precluded analysis by intended route of delivery or the assessment of other pertinent neonatal or maternal morbidities. Lastly, most data on the impact of route of delivery on outcomes of triplet gestations are >20 years old and therefore may not reflect current obstetric or neonatal practice patterns.

Therefore, the primary objectives of this study are to determine the likelihood of success in an attempted vaginal delivery and assess maternal and neonatal outcomes of attempted vaginal vs planned cesarean delivery of triplets using a large, multiinstitution obstetric cohort.

Materials and Methods

We performed a secondary analysis of data from the Consortium on Safe Labor, a retrospective observational study conducted by the Eunice Kennedy Shriver National Institute of Child Health and Human Development to characterize current labor and delivery clinical practice and outcomes. Twelve clinical centers including 19 hospitals participated in the Consortium on Safe Labor, which represented all 9 American Congress of Obstetricians and Gynecologists districts that existed at the time of the study. Data were collected on 228,668 deliveries occurring from 2002 through 2008 with the majority of births (87%) occurring from 2005 through 2007. All deliveries at a gestational age >23 weeks were included in the data set. Participating institutions submitted detailed information from patient electronic medical records, including maternal demographic characteristics; maternal medical, reproductive, and antenatal histories; labor and delivery data; and postpartum maternal and neonatal outcome data. Complete details of the study methodology have been previously published. As a secondary analysis of deidentified data, this study was considered exempt by the MetroHealth Medical Center institutional review board.

The following methodology was used to generate our study cohort. The Consortium on Safe Labor data set contains a predefined variable for fetal number, allowing for the identification of all triplet gestations. We restricted our cohort to women delivering at a gestational age ≥28 weeks given the unlikely recommendation for planned vaginal delivery earlier than this gestational age. To arrive at our final cohort, women with a history of cesarean delivery and pregnancies complicated by congenital or chromosomal anomalies, antenatal fetal death of ≥1 fetuses, and twin-twin transfusion syndrome were then excluded. Exclusion criteria were applied at the level of the neonate (and not the level of the pregnancy) to ensure exclusion of all pregnancies where any single member of a triplet set met an exclusion criterion. Only complete triplet sets were retained in the final cohort. Given that estimated fetal weight was not available for the majority of the cohort (213 of 240 fetuses), the exclusion of pregnancies ineligible for vaginal delivery based on this criterion was not feasible.

Women in our triplet pregnancy cohort were then divided into 1 of 2 modes of delivery: attempted vaginal delivery or planned cesarean delivery. We used previously defined methodology from publications from the Consortium on Safe Labor to generate these 2 study groups. In short, this methodology allowed for the identification of labor type (spontaneous vs induced) and route of delivery (vaginal, operative vaginal, prelabor cesarean, or intrapartum cesarean). Various indications for labor induction or intrapartum cesarean delivery are distinctly coded in the Consortium on Safe Labor data set. For this study, the planned cesarean delivery group included only women undergoing cesarean delivery and excluded anyone coded as having an induction or augmentation of labor, episiotomy, perineal laceration, recorded time of labor onset or complete dilation, or cesarean delivery for an intrapartum indication (eg, failed operative vaginal delivery or labor dystocia). Women delivering by cesarean delivery with the coded indications of “elective” and “fetal malpresentation” were also included in the planned cesarean delivery group. As such, this group best approximates women undergoing prelabor, planned cesarean delivery. The attempted vaginal delivery group was defined to include all women with evidence of spontaneous, induced, or augmented labor and excluded anyone delivering by planned or prelabor cesarean (as previously defined). Therefore, this group best approximates women laboring with the assumption of a planned vaginal birth.

The primary outcomes for this study were maternal and neonatal morbidity. Maternal morbidity included the individual outcomes of infection during delivery hospitalization (composite of chorioamnionitis, endometritis, wound separation, and wound infection), blood transfusion, or transfer to the intensive care unit. Primary neonatal outcome measures included the individual outcomes of neonatal asphyxia and mechanical ventilation as well as composite neonatal morbidity, consisting of ≥1 of the following: birth injury, 5-minute Apgar <4, arterial pH <7.0 or base excess <–12.0, neonatal asphyxia, or neonatal death. For the assessment of patient characteristics, baseline data, and maternal outcome measures, the unit of analysis was the pregnancy. Continuous variables were compared using the unpaired t test and categorical variables were compared using χ ² or Fisher exact test where appropriate. Given the inherent clustering present in triplet births, outcomes between fetuses are correlated and not independent. Therefore, the analysis of binary neonatal outcomes was performed using Poisson regression with clustering to account for correlation between neonates within a triplet pregnancy. Controlling for confounding factors was performed as outcome rates allowed and results are reported as adjusted incidence rate ratios. Factors included in the regression model were chosen based on biologic plausibility (gestational age, birth order) and the results of univariable analysis (antenatal corticosteroids).

Lastly, to further minimize confounding by gestational age and assess the validity of the attempted vaginal delivery group, a sensitivity analysis was performed in the subcohort of women at gestational age ≥34 weeks. In this sensitivity analysis, the attempted vaginal delivery group was restricted to include only women with evidence of induction or augmentation or labor (and thereby eliminate women presenting in spontaneous labor with an early, and likely planned, cesarean delivery). The same statistical methodology was employed to assess maternal and neonatal outcome measures in the subcohort. Statistical significance was set at P < .05. All analysis was performed using software (Stata 13.1; StataCorp, College Station, TX).

Materials and Methods

We performed a secondary analysis of data from the Consortium on Safe Labor, a retrospective observational study conducted by the Eunice Kennedy Shriver National Institute of Child Health and Human Development to characterize current labor and delivery clinical practice and outcomes. Twelve clinical centers including 19 hospitals participated in the Consortium on Safe Labor, which represented all 9 American Congress of Obstetricians and Gynecologists districts that existed at the time of the study. Data were collected on 228,668 deliveries occurring from 2002 through 2008 with the majority of births (87%) occurring from 2005 through 2007. All deliveries at a gestational age >23 weeks were included in the data set. Participating institutions submitted detailed information from patient electronic medical records, including maternal demographic characteristics; maternal medical, reproductive, and antenatal histories; labor and delivery data; and postpartum maternal and neonatal outcome data. Complete details of the study methodology have been previously published. As a secondary analysis of deidentified data, this study was considered exempt by the MetroHealth Medical Center institutional review board.

The following methodology was used to generate our study cohort. The Consortium on Safe Labor data set contains a predefined variable for fetal number, allowing for the identification of all triplet gestations. We restricted our cohort to women delivering at a gestational age ≥28 weeks given the unlikely recommendation for planned vaginal delivery earlier than this gestational age. To arrive at our final cohort, women with a history of cesarean delivery and pregnancies complicated by congenital or chromosomal anomalies, antenatal fetal death of ≥1 fetuses, and twin-twin transfusion syndrome were then excluded. Exclusion criteria were applied at the level of the neonate (and not the level of the pregnancy) to ensure exclusion of all pregnancies where any single member of a triplet set met an exclusion criterion. Only complete triplet sets were retained in the final cohort. Given that estimated fetal weight was not available for the majority of the cohort (213 of 240 fetuses), the exclusion of pregnancies ineligible for vaginal delivery based on this criterion was not feasible.

Women in our triplet pregnancy cohort were then divided into 1 of 2 modes of delivery: attempted vaginal delivery or planned cesarean delivery. We used previously defined methodology from publications from the Consortium on Safe Labor to generate these 2 study groups. In short, this methodology allowed for the identification of labor type (spontaneous vs induced) and route of delivery (vaginal, operative vaginal, prelabor cesarean, or intrapartum cesarean). Various indications for labor induction or intrapartum cesarean delivery are distinctly coded in the Consortium on Safe Labor data set. For this study, the planned cesarean delivery group included only women undergoing cesarean delivery and excluded anyone coded as having an induction or augmentation of labor, episiotomy, perineal laceration, recorded time of labor onset or complete dilation, or cesarean delivery for an intrapartum indication (eg, failed operative vaginal delivery or labor dystocia). Women delivering by cesarean delivery with the coded indications of “elective” and “fetal malpresentation” were also included in the planned cesarean delivery group. As such, this group best approximates women undergoing prelabor, planned cesarean delivery. The attempted vaginal delivery group was defined to include all women with evidence of spontaneous, induced, or augmented labor and excluded anyone delivering by planned or prelabor cesarean (as previously defined). Therefore, this group best approximates women laboring with the assumption of a planned vaginal birth.

The primary outcomes for this study were maternal and neonatal morbidity. Maternal morbidity included the individual outcomes of infection during delivery hospitalization (composite of chorioamnionitis, endometritis, wound separation, and wound infection), blood transfusion, or transfer to the intensive care unit. Primary neonatal outcome measures included the individual outcomes of neonatal asphyxia and mechanical ventilation as well as composite neonatal morbidity, consisting of ≥1 of the following: birth injury, 5-minute Apgar <4, arterial pH <7.0 or base excess <–12.0, neonatal asphyxia, or neonatal death. For the assessment of patient characteristics, baseline data, and maternal outcome measures, the unit of analysis was the pregnancy. Continuous variables were compared using the unpaired t test and categorical variables were compared using χ ² or Fisher exact test where appropriate. Given the inherent clustering present in triplet births, outcomes between fetuses are correlated and not independent. Therefore, the analysis of binary neonatal outcomes was performed using Poisson regression with clustering to account for correlation between neonates within a triplet pregnancy. Controlling for confounding factors was performed as outcome rates allowed and results are reported as adjusted incidence rate ratios. Factors included in the regression model were chosen based on biologic plausibility (gestational age, birth order) and the results of univariable analysis (antenatal corticosteroids).

Lastly, to further minimize confounding by gestational age and assess the validity of the attempted vaginal delivery group, a sensitivity analysis was performed in the subcohort of women at gestational age ≥34 weeks. In this sensitivity analysis, the attempted vaginal delivery group was restricted to include only women with evidence of induction or augmentation or labor (and thereby eliminate women presenting in spontaneous labor with an early, and likely planned, cesarean delivery). The same statistical methodology was employed to assess maternal and neonatal outcome measures in the subcohort. Statistical significance was set at P < .05. All analysis was performed using software (Stata 13.1; StataCorp, College Station, TX).