The incidence of twin pregnancy has risen over the past 20 years, due mainly to infertility treatments and older maternal age.1–3 In 2016, the twin pregnancy rate was around 1.8% in France and 1.7% in the United States.1,4 Both neonatal and maternal complications are more common with twin pregnancies than singleton pregnancies and include premature birth, hypertensive disorders of pregnancy, obstetric hemorrhage and abnormal placentation.5–9 However, our knowledge of the association of twin pregnancies with maternal severe health effects and organ dysfunction is incomplete. Most of the few previous studies reporting an increased risk of severe maternal outcomes in twin pregnancies have notable methodologic limitations, including their analysis of retrospective or administrative databases, failure to control for confounders, and use of an outcome definition that is limited to selected obstetric complications and fails to take the woman's health into account.10–14 A recent study conducted by the World Health Organization (WHO) mostly in low- or middle-income countries reports a higher risk of “maternal near-miss” events in twin pregnancies, but the generalizability of its results for high-income countries remains controversial given the contrasts in practices and organization of care.12,15 In addition, it seems important to characterize intermediate factors that might lie on the causal pathway from twin pregnancy to severe maternal morbidity. Literature about these potential intermediate factors is even more limited, and route of delivery might be a major candidate for such a factor because cesarean deliveries are more frequent in twin pregnancies and have been associated with a higher risk of severe maternal morbidity than vaginal births.16–19 Our objective was to investigate the association between twin pregnancy and severe acute maternal morbidity, overall and by timing (antepartum or intrapartum or postpartum) and underlying causal condition. A secondary objective was to explore the contribution of cesarean delivery as an intermediate factor in this association.
This population-based, cohort-nested, case-control analysis used data from the EPIMOMS study, a prospective study specifically designed to study severe acute maternal morbidity that was conducted in six French regions from May 2012 through November 2013.20 The EPIMOMS study includes a standardized definition of severe acute maternal morbidity that was developed through a national Delphi formal expert consensus process and was intended to characterize maternal complications with severe health effects and organ dysfunction. The EPIMOMS definition of severe acute maternal morbidity combines diagnoses (major obstetric bleeding, eclampsia, severe preeclampsia, pulmonary embolism, stroke, and psychiatric disorder), organ dysfunctions (cardiovascular, respiratory, renal, neurologic, hepatic, and hematologic), and interventions (intensive care admission and laparotomy after delivery) (definition detailed in Appendix 2, available online at http://links.lww.com/AOG/B362).
Recruitment took place over a 1-year period in each region at hospitals, with 119 maternity units and 136 intensive care units participating. They comprised 182,309 deliveries, approximately 20% of those in France during the study period, and the characteristics of both parturients and hospitals were similar to the national profile.21 Although all deliveries were counted, only minimal information was available for all, including multiple or singleton status. For all those with severe acute maternal morbidity (case group) and a random sample of those without (control group), specific detailed information was collected. This design has the assets of a cohort (in particular no selection bias owing to a lack of comparability of women in the case and control groups, because they are selected from the exactly same source population) while minimizing the number of women included and the related costs.
All women who experienced a severe acute maternal morbidity event from 22 weeks of gestation and up to 42 days postpartum were prospectively identified and included in EPIMOMS. The completeness of case identification was controlled by review of delivery logbooks, hospital discharge data, and laboratory files. Concomitantly, at the end of the inclusion period, a 2% unmatched random sample of women who gave birth without severe acute maternal morbidity in the same regions during the same time period was selected from the maternity units' birth registers as the control group. To show an odds ratio (OR) of at least 1.5 but no more than 0.67 for a factor with a prevalence of 5% or more among women in the control group with α=0.05 and 1 − β=0.9, 3,600 women (or 2% of the cohort) were needed for the control group. For each maternity unit, 2% of their delivering patients were selected for the control group. Women in the control group were those without severe acute maternal morbidity; this definition did not exclude nonsevere maternal morbidity. The sampling method was to randomly select a number between one and 50, and then to select one every 50 deliveries during the 1-year study period in each unit; in case this process selected a woman with severe acute maternal morbidity, the delivery just before was selected as a control.
Data about women's social and demographic characteristics, preexisting medical and obstetric conditions, characteristics and complications of the current pregnancy and delivery, and details about the course and management of the severe acute maternal morbidity were collected from medical records in which those items were prospectively recorded as they occurred. The data collection method was similar for women with severe acute maternal morbidity and those in the control group and in all regions. Information was extracted from paper charts by research midwives trained for this study and was entered in a specific electronic report form developed for the study. In addition, for women with severe acute maternal morbidity, the criteria and causes of severe acute maternal morbidity were entered by the clinician in charge. Because the severe acute maternal morbidity case criteria does not necessarily correspond directly to its cause, we also recorded the causal condition and clinical timing (antepartum, intrapartum or postpartum), as recorded prospectively by the clinician in charge.
For this study, we excluded triplet pregnancies because this subgroup is both likely to be at specific risk of severe acute maternal morbidity and is too small to be distinguished in the analysis. We further excluded 30 women with significant obstetric data missing, most likely because they developed severe acute maternal morbidity owing to a nonobstetric condition (trauma or psychiatric disorder) early in pregnancy and were then lost to follow-up (Fig. 1).
The primary outcome was severe acute maternal morbidity, as a binary variable. Women who experienced more than one such event were counted once in this analysis; however, we described the number of severe acute maternal morbidity events per woman. To isolate the most severe fraction of severe acute maternal morbidity cases, a secondary outcome was defined to include maternal near-misses, as defined by WHO15 (see criteria in Box 1). Given the extreme rarity of maternal deaths, they were included in this more severe near-miss outcome and not isolated as a specific outcome; for brevity and clarity, this group will be referred to as the “near-miss” group in this article.
World Health Organization Maternal Near-Miss Criteria
- Cardiac arrest (absence of pulse or heart beat and loss of consciousness)
- Use of continuous vasoactive drugs
- Cardiopulmonary resuscitation
- Severe hypoperfusion (lactate greater than 5 mmol/L or greater than 45 mg/dL)
- Severe acidosis (pH less than 7.1)
- Acute cyanosis
- Severe tachypnea (respiratory rate greater than 40 breaths/min)
- Severe bradypnea (respiratory rate less than 6 breaths/min)
- Intubation and ventilation (for 60 min or longer*) not related to anesthesia
- Severe hypoxemia (O2 saturation less than 90% (for 60 min or longer*) or PaO2/FiO2 less than 200 mm Hg)
- Oliguria nonresponsive to fluids or diuretics
- Dialysis for acute renal failure
- Severe acute azotemia (creatinine 300 micromole/L or greater or 3.5 mg/dL or greater)
- Failure to form clots
- Massive transfusion of blood or red cells (5 units or more)
- Severe acute thrombocytopenia (less than 50,000 platelets/mL)
- Jaundice in the presence of preeclampsia
- Severe acute hyperbilirubinemia (bilirubin greater than 100 micromole/L or greater than 6.0 mg/dL)
- Prolonged unconsciousness (lasting 12 hours or longer*) or coma (including metabolic coma)
- Uncontrollable fit or epileptic status
- Total paralysis
- Uterine hemorrhage or infection leading to hysterectomy
Data from Say L, Souza JP, Pattinson RC, WHO working group on Maternal Mortality Morbidity classifications. Maternal near miss--towards a standard tool for monitoring quality of maternal health care. Best Pract Res Clin Obstet Gynaecol 2009;23):287–96.
* Data not available in the EPIMOMS study: gasping, duration of intubation and ventilation, duration of hypoxemia and duration of prolonged unconsciousness.
The analysis of the association between twin pregnancy and severe acute maternal morbidity considered the covariables listed in Box 2.
Covariables Considered in the Analysis of the Association Between Twin Pregnancy and Severe Acute Maternal Morbidity
- Age (continuous variable, in years)
- Country or birth (France, other Europe, North Africa, sub-Saharan Africa, others)
- Single mother
- Body mass index ([calculated as weight in kilograms divided by height in meters squared] underweight [less than 18.5], normal [18.5–24.9], overweight [25.0–29.9], obese [30 or higher])
- Preexisting medical condition (binary variable defined as at least one of the following conditions: cardiovascular disease, respiratory disease, stroke, bleeding disorder, thromboembolic disease, hemoglobinopathy, thyroid dysfunction, autoimmune disease, gastrointestinal disease, nephropathy, cancer, epilepsy, or psychiatric disorder)
- History of abdominopelvic surgery
- Parity (nulliparous or parous)
- Previous cesarean delivery
- Previous obstetric hemorrhage
- Previous hypertensive disorder in pregnancy
- Conception by in vitro fertilization
- Abnormal placentation
- Hypertensive disorder (eg, gestational hypertension or nonsevere preeclampsia)
- Third-trimester anemia, defined as a hemoglobin level less than 11 g/dL
- Gestational age at delivery
- Mode of delivery (spontaneous vaginal, operative vaginal, antepartum cesarean, intrapartum cesarean)
Near-miss status, timing of severe acute maternal morbidity, and underlying causal conditions were described among the women in the case group according to twin or singleton pregnancy status. For women who had several severe acute maternal morbidity events, the timing of the first event was considered.
We compared the characteristics of women in the case and control groups based on Pearson's χ2 test or Fisher exact test for categorical variables, and Student t-test or Wilcoxon-Mann-Whitney test for quantitative variables, as appropriate. Statistical significance was set at a two-tailed value of P<.05.
The association between twin pregnancy and severe acute maternal morbidity was analyzed with multivariable multilevel logistic regression to take into account the data's hierarchical structure (women within maternity units). Causal assumptions between twin pregnancy, severe acute maternal morbidity, and covariates were represented with a directed acyclic graph to depict the exposure-outcome relations with confounding and intermediate factors (Appendix 3, available online at http://links.lww.com/AOG/B362). The directed acyclic graph, thus, helped to select variables that are confounders (ie, variables associated with both the exposure, which is twin pregnancy, and the outcome of severe acute maternal morbidity, and not on the causal pathway between twin pregnancy and severe acute maternal morbidity) and those that do not qualify as confounders (especially intermediate factors).22 The main regression model included only strict confounders: maternal age, maternal country of birth, parity, and conception by in vitro fertilization (IVF). Because the relation between severe acute maternal morbidity and maternal age (continuous variable) was not linear, this variable was transformed into second degree fractional polynomials. The only potential interaction considered clinically relevant was between twin pregnancy and maternal age, and it was not significant.
A sensitivity analysis used a regression model that included, in addition to the confounders of the main model, other variables that are risk factors for severe acute maternal morbidity but are not associated with twin pregnancy (ie, single mother [used as a proxy for social vulnerability], body mass index, previous cesarean delivery, preexisting medical condition, previous hypertensive disorder in pregnancy, and previous obstetric hemorrhage).
The same analysis strategy was used to assess the associations between twin pregnancy and each of the following specific categories of severe acute maternal morbidity: maternal near-miss events, the timing of severe acute maternal morbidity occurrence (antepartum and intrapartum or postpartum), and the underlying causal conditions (severe obstetric hemorrhage, severe hypertensive complications, and all other underlying causal conditions).
The proportion of women with missing data for any covariate included in the main multivariable model ranged from 0% to 13.7%. There were 2,227 (89.1%) women in the case group and 3,048 (83.5%) in the control group with no missing data for covariates included in the main multivariable model. Characteristics of the women with full data were similar to those with missing data (Appendix 4, available online at http://links.lww.com/AOG/B362). We used multiple imputation chained equations according to Rubin's rules to impute missing data (20 data sets imputed). The results of the univariable and multivariable analysis are presented with the imputed data. We also performed an analysis with the nonimputed data.
A secondary analysis estimated the contribution of cesarean delivery (vs vaginal delivery) as an intermediate factor in the association between twin pregnancy and severe acute maternal morbidity by path analysis.22 The first step was to select a relevant population to minimize the indication bias between cesarean delivery and severe acute maternal morbidity, that is, the fact that some clinical situations will influence mode of delivery, specifically the indication for cesarean delivery, and the risk of severe acute maternal morbidity, and can thus lead to incorrectly attributing a risk present before the procedure to the cesarean delivery. We therefore excluded women with antepartum severe acute maternal morbidity (n=585); women in situations in which one or the other mode of delivery is used systematically, either cesarean (for fetal transverse position or unknown position [n=59] or for at least two previous cesarean deliveries [n=97]) or vaginal (home births [n=10], medical termination of pregnancy [n=7], in utero and intrapartum fetal deaths [n=40]); women with obstetric conditions that developed during pregnancy, were symptomatic and present before labor but without severe acute maternal morbidity at this stage, and responsible for a postpartum severe acute maternal morbidity (for example gestational hypertension or placenta praevia) (n=233); women with severe acute maternal morbidity during labor that required an emergency cesarean delivery (n=6); and women with unknown mode of delivery (n=1). These exclusions left 1,462 women with intrapartum or postpartum severe acute maternal morbidity including 1,253 (85.7%) without missing data for the variables included in this secondary model (see below), who comprised the case group for this secondary analysis. Applying similar exclusion criteria to the control group resulted in 3,464 women without severe acute maternal morbidity, including 2,816 (81.3%) without missing data, who comprised the control group for this secondary analysis.
Following the method proposed by Erikson et al and generalized by Buis,23,24 the total effect of twin pregnancy on severe acute maternal morbidity was decomposed into an indirect effect mediated by cesarean delivery and a direct (or residual) effect not mediated by cesarean delivery. This method of analysis is based on a counterfactual approach to causality and confounding that can be formulated as the answer to the following question: “What would the risk of severe acute maternal morbidity associated with twin pregnancy be if twin pregnancies had the same probability of cesarean delivery as singleton pregnancies?” The indirect effect of twin pregnancy on the risk of severe acute maternal morbidity, mediated by cesarean delivery, was calculated from the regression coefficients obtained through logistic regression models, adjusted for the same covariates as the main model, but also for confounders in the association between mode of delivery and severe acute maternal morbidity (Appendix 3, http://links.lww.com/AOG/B362). This indirect effect was expressed as a percentage of the total effect.
At the conventional two-tailed significance level of α=0.05, with a twin pregnancy prevalence of 1.6%, the available number of women in the case and control groups provided a statistical power of 80% to show an OR greater than or equal to 1.7 quantifying the association between twin pregnancy and severe acute maternal morbidity. STATA 13 software was used for the descriptive and multivariable analyses.
The Commission Nationale de l’Informatique et des Libertés (CNIL, no 912210), the French data protection agency, approved the EPIMOMS study.
The total cohort included 182,309 women, of whom 1.4% (95% CI 1.3–1.4%) (n=2,540) had severe acute maternal morbidity. Among 3,202 twin pregnancies, severe acute maternal morbidity occurred in 197 (6.2%, 95% CI 5.3–7.1%) and in 2,303 of the 179,107 singleton pregnancies (1.3%, 95% CI 1.2–1.3%). Our analysis included 2,500 women with severe acute maternal morbidity (case group) and 3,650 women without severe acute maternal morbidity (control group). Most cases of severe acute maternal morbidity occurred intrapartum or postpartum (76.6%, n=1,915/2,500). The two most common underlying causal conditions of severe acute maternal morbidity were severe obstetric hemorrhage (65.6%, n=1,641/2,500) and severe hypertensive complications (19.5%, n=488/2,500) (Table 1). Maternal near-misses accounted for 37.4% of severe acute maternal morbidity cases (n=934/2,500), including 13 maternal deaths, all of which occurred in singleton pregnancies. The distribution of the timing, underlying causal conditions and severity of severe acute maternal morbidity did not differ significantly in twin and singleton pregnancies (Appendix 5, available online at http://links.lww.com/AOG/B362).
Among women in the case group, 7.9% (197/2,500) had twin pregnancies; among women in the control group, 1.6% (59/3,650) had twin pregnancies (unadjusted OR 4.7, 95% CI 3.5–6.4) (Tables 2 and 3). Compared with women in the control group, women who experienced a severe acute maternal morbidity event were older; they were more often born in sub-Saharan Africa and “other” countries, nulliparous, and pregnant by IVF. They also had a higher frequency of preexisting medical conditions, history of abdominopelvic surgery, obstetric complications, previous cesarean deliveries, and preterm and cesarean deliveries (Table 2).
In the multivariable analysis, compared with singleton pregnancies, twin pregnancies were at higher risk of severe acute maternal morbidity (adjusted OR 4.2, 95% CI 3.1–5.8), both antepartum (adjusted OR 4.1, 95% CI 2.5–6.6) and intrapartum or postpartum (adjusted OR 4.2, 95% CI 3.1–5.9), from severe obstetric hemorrhage (adjusted OR 4.8, 95% CI 3.5–6.7), severe hypertensive complications (adjusted OR 4.8, 95% CI 2.9–8.0), and other underlying causal conditions (adjusted OR 3.0, 95% CI 1.7–5.2), and of near-misses (adjusted OR 5.1, 95% CI 3.5–7.3) (Table 3). The sensitivity analysis adjusted for strict confounders and other risk factors for severe acute maternal morbidity yielded similar results (Table 3), as did the analyses with nonimputed data (results not shown).
In the population selected for the secondary analysis, the cesarean delivery rate among women in the control group was 57.7% for twin pregnancies and 17.6% for singleton pregnancies (P<.001); among women in the case group, it was 71.7% and 34.1% (P<.001), respectively. Table 4 shows the results of the path analysis to assess the contribution of cesarean delivery as an intermediate factor in the relation between twin pregnancy and intrapartum or postpartum severe acute maternal morbidity. In this secondary analysis, the association between twin pregnancy and intrapartum or postpartum severe acute maternal morbidity was similar to that in the main analysis (crude OR 5.9, 95% CI 4.2–8.2; and adjusted OR 5.2, 95% CI 3.5–7.7). Therefore, 20.6% (95% CI 12.9–28.2%) of the total risk of intrapartum or postpartum severe acute maternal morbidity associated with twin pregnancy was mediated by cesarean delivery.
We found a strong association between twin pregnancy and severe acute maternal morbidity: The risk was four times higher than for singleton pregnancy after controlling for identified confounders. This excess risk was also found for maternal near-misses and before, during or after delivery. Path analysis taking into account the potential indication bias suggests that about one fifth of the association between twin pregnancy and intrapartum or postpartum severe acute maternal morbidity was mediated by cesarean delivery; in other words, if twin pregnancies had the same probability of cesarean delivery as singleton pregnancies, the association found between twin pregnancy and intrapartum or postpartum severe acute maternal morbidity would be reduced by one fifth.
This population-based design allowed us to take into account the diversity of women and practices that can affect severe maternal morbidity. The national Delphi-Rand formalized process produced a consensual and comprehensive definition of severe acute maternal morbidity based on specific criteria for severe health effects and organ dysfunction. The prospective identification of women for the case group was as exhaustive as possible; all severe acute maternal morbidity events corresponding to this precise definition were identified during the period and checked afterwards. The detailed data of maternal and severe acute maternal morbidity characteristics made it possible to apply the WHO definition of maternal near-miss to our population and to distinguish severe acute maternal morbidity by clinical timing, which has practical implications, and underlying causal condition, as well as to consider a large number of confounders. Path analysis furnished additional information about the potential intermediate role of cesarean delivery in the association between twin pregnancy and intrapartum or postpartum severe acute maternal morbidity. Our special focus on controlling confounding by indication for this secondary analysis minimized this bias, which often impairs analyses of the adverse consequences of cesarean deliveries.
The external validity of this study may be limited by the participating population identified from six French regions. Although this was not a random sample of French parturients, the characteristics of women and maternity units in this large source population (one fifth of national deliveries) were similar to the national profile.21 Nevertheless, obstetric practices and the management of twin pregnancies vary across countries, and these differences may limit the generalizability of our results.
Without distinguishing the level of severity of maternal complications and based on international retrospective hospital data, Conde-Agudelo et al10 reported a twofold adjusted (for maternal age and parity) risk of hypertensive disorder in pregnancy and postpartum hemorrhage for twin pregnancies. Studying 1995–2000 birth certificates in the United States, Luke et al found a significant crude association between twin pregnancies and adverse obstetric outcomes, but did not distinguish severe maternal morbidity.11 The analysis by Walker et al13 of 1984–2000 Canadian data showed a higher risk of adverse obstetric outcomes in twin pregnancies, without adjustment for potential confounders. The validity of the data provided by hospital-based retrospective studies and administrative databases remains questionable, and to our knowledge, only a few observational prospective studies were designed to study severe maternal morbidity. In a cohort of 2,552 cases of severe maternal morbid events, including 202 twin pregnancies in 2004–2006, Witteveen et al14 in the Netherlands showed a significant association between twin pregnancy and severe maternal morbidity (OR 4.3, 95% CI 3.7–5.0), without adjustment for potential confounding factors. Furthermore, their definition of severe maternal morbidity was restricted to certain components (intensive care unit admission, uterine rupture, eclampsia, severe hemorrhage). In a prospective international study conducted by WHO in 2010–2011, twin pregnancy was significantly associated with maternal near-miss (OR adjusted for the country of birth 3.03, 95% CI 2.39–3.85).12 After adjustment for maternal age, parity, medical conditions, obstetric complications, and mode of delivery, but restricted to low- and middle-income countries, Vogel et al25 found an adjusted OR of severe adverse maternal outcome (including death) of 1.85 (95% CI 1.60–2.14) in twin pregnancies. We were able to apply the WHO maternal near-miss definition to the women in the case group and found crude and adjusted ORs of maternal near-miss or death in twin pregnancies of 5.7 (95% CI 4.0–8.1) and 5.1 (95% CI 3.5–7.3), respectively. These differences may be explained by the dominant participation of medium to low-income countries, with their higher levels of risk, including for the reference group of singleton pregnancies.
Results of the analysis of the risk of severe acute maternal morbidity by timing and by underlying causal condition are informative for clinical practice. Twin delivery is known to be at higher risk of postpartum hemorrhage, and our analysis confirmed and quantified this association for the most severe hemorrhages and after adequate controlling for confounders. However, we showed that the risk of severe acute maternal morbidity in women with twin pregnancies is also markedly higher in the antepartum period and for underlying causal conditions other than severe obstetric hemorrhage. These results should encourage obstetricians to provide women carrying twins with close medical monitoring during prenatal care—an example may be the detection of antepartum anemia with iron supplementation to prevent severe maternal anemia postdelivery—and to plan the delivery in a maternity unit appropriate for these maternal risks, in particular, units with on-site extensive access to blood products, adult intensive care unit, and surgical expertise. This finding also underlines the potential consequences of infertility treatments, including multiple embryo transfer in IVF, an important provider of twin pregnancies2 (one quarter of twin pregnancies in France in 20103).
We found that about one fifth of the association between twin pregnancy and intrapartum or postpartum severe acute maternal morbidity may be mediated by cesarean delivery. Although we did not differentiate cesarean delivery by timing in this analysis, this finding provides further support to the recommendation to limit the use of cesarean for twin deliveries to situations where it is clearly medically indicated.26 In our study, more than half of women in the control group with twin pregnancies and about three quarters of women in the case group with twin pregnancies delivered by cesarean. Increasing the rate of vaginal deliveries in twin pregnancies, unless there is an obstetric or fetal indication otherwise, may decrease the rate of severe acute maternal morbidity in these women.
1. Blondel B, Coulm B, Bonnet C, Goffinet F, Le Ray C; National Coordination Group of the National Perinatal Surveys. Trends in perinatal health in metropolitan France from 1995 to 2016: results from the French National Perinatal Surveys. J Gynecol Obstet Hum Reprod 2017;46:701–13.
2. Kulkarni AD, Jamieson DJ, Jones HW Jr, Kissin DM, Gallo MF, Macaluso M, et al. Fertility treatments and multiple births in the United States. N Engl J Med 2013;369:2218–25.
3. Prunet C, Goffinet F, Blondel B. Medical care and perinatal health in twin pregnancies: situation in 2010 and recent trends in France [French]. J Gynecol Obstet Biol Reprod (Paris) 2015;44:184–93.
4. Martin JA, Hamilton BE, Osterman MJK, Driscoll AK, Drake P. Births: final data for 2016. Natl Vital Stat Rep 2018;67:1–55.
5. Gyamfi C, Stone J, Eddleman KA. Maternal complications of multifetal pregnancy. Clin Perinatol 2005;32:431–42, vii.
6. Senat MV, Ancel PY, Bouvier-Colle MH, Breart G. How does multiple pregnancy affect maternal mortality and morbidity? Clin Obstet Gynecol 1998;41:78–83.
7. Young BC, Wylie BJ. Effects of twin gestation on maternal morbidity. Semin Perinatol 2012;36:162–8.
8. Heino A, Gissler M, Hindori-Mohangoo AD, Blondel B, Klungsoyr K, Verdenik I, et al. Variations in multiple birth rates and impact on perinatal outcomes in Europe. PLoS One 2016;11:e0149252.
9. Rao A, Sairam S, Shehata H. Obstetric complications of twin pregnancies. Best Pract Res Clin Obstet Gynaecol 2004;18:557-76.
10. Conde-Agudelo A, Belizan JM, Lindmark G. Maternal morbidity and mortality associated with multiple gestations. Obstetrics Gynecol 2000;95:899–904.
11. Luke B, Brown MB. Contemporary risks of maternal morbidity and adverse outcomes with increasing maternal age and plurality. Fertil Steril 2007;88:283–93.
12. Santana DS, Cecatti JG, Surita FG, Silveira C, Costa ML, Souza JP, et al. Twin pregnancy and severe maternal outcomes: the World Health Organization multicountry survey on maternal and newborn health. Obstetrics Gynecol 2016;127:631–41.
13. Walker MC, Murphy KE, Pan S, Yang Q, Wen SW. Adverse maternal outcomes in multifetal pregnancies. Int J Obstet Gynaecol 2004;111:1294–6.
14. Witteveen T, Van Den Akker T, Zwart JJ, Bloemenkamp KW, Van Roosmalen J. Severe acute maternal morbidity in multiple pregnancies: a nationwide cohort study. Am J Obstet Gynecol 2016;214:641.e1–10.
15. Say L, Souza JP, Pattinson RC; WHO Working Group on Maternal Mortality Morbidity Classifications. Maternal near miss—towards a standard tool for monitoring quality of maternal health care. Best Pract Res Clin Obstet Gynaecol 2009;23:287–96.
16. Barrett JF, Hannah ME, Hutton EK, Willan AR, Allen AC, Armson BA, et al. A randomized trial of planned cesarean or vaginal delivery for twin pregnancy. N Engl J Med 2013;369:1295–305.
17. Hutton EK, Hannah ME, Ross S, Joseph KS, Ohlsson A, Asztalos EV, et al. Maternal outcomes at 3 months after planned caesarean section versus planned vaginal birth for twin pregnancies in the Twin Birth Study: a randomised controlled trial. Int J Obstet Gynaecol 2015;122:1653–62.
18. Liu S, Liston RM, Joseph KS, Heaman M, Sauve R, Kramer MS, et al. Maternal mortality and severe morbidity associated with low-risk planned cesarean delivery versus planned vaginal delivery at term. CMAJ 2007;176:455–60.
19. Villar J, Carroli G, Zavaleta N, Donner A, Wojdyla D, Faundes A, et al. Maternal and neonatal individual risks and benefits associated with caesarean delivery: multicentre prospective study. BMJ 2007;335:1025.
20. Siddiqui A, Azria E, Howell EA, Deneux-Tharaux C; EPIMOMS Study Group. Associations between maternal obesity and severe maternal morbidity: findings from the French EPIMOMS population-based study. Paediatric Perinatal Epidemiol 2019;33:7–16.
21. Blondel B, Lelong N, Kermarrec M, Goffinet F; National Coordination Group of the National Perinatal Surveys. Trends in perinatal health in France from 1995 to 2010. Results from the French National Perinatal Surveys. J Gynecol Obstet Biol Reprod (Paris) 2012;41:e1–15.
22. Ananth CV, Schisterman EF. Confounding, causality, and confusion: the role of intermediate variables in interpreting observational studies in obstetrics. Am J Obstet Gynecol 2017;217:167–75.
23. Buis ML. Direct and indirect effects in a logit model. Stata J 2010;10:11–29.
24. Erikson R, Goldthorpe JH, Jackson M, Yaish M, Cox DR. On class differentials in educational attainment. Proc Natl Acad Sci U S A 2005;102:9730–3.
25. Vogel JP, Torloni MR, Seuc A, Betran AP, Widmer M, Souza JP, et al. Maternal and perinatal outcomes of twin pregnancy in 23 low- and middle-income countries. PLoS One 2013;8:e70549.
26. Korb D, Deneux-Tharaux C, Seco A, Goffinet F, Schmitz T. JUmeaux MODe d'Accouchement Study Group, et al. Risk of severe acute maternal morbidity according to planned mode of delivery in twin pregnancies. Obstetrics Gynecol 2018;132:647–55.