Mode and timing of twin delivery and perinatal outcomes in low- and middle-income countries: a secondary analysis of the WHO Multicountry Survey on Maternal and Newborn Health

Authors

  • T Ganchimeg,

    1. Department of Health Policy, National Center for Child Health and Development, Tokyo, Japan
    Search for more papers by this author
  • N Morisaki,

    Corresponding author
    1. Department of Health Policy, National Center for Child Health and Development, Tokyo, Japan
    2. Department of Paediatrics, Graduate School of Medicine, University of Tokyo, Tokyo, Japan
    • Correspondence: Dr N Morisaki, Department of Health Policy, National Centre for Child Health and Development, 10-1-2 Okura, Setagaya-ku, Tokyo, 157-8535, Japan. Email morisaki-n@ncchd.go.jp

    Search for more papers by this author
  • JP Vogel,

    1. Department of Reproductive Health and Research, UNDP/UNFPA/UNICEF/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction (HRP), World Health Organization, Geneva, Switzerland
    2. School of Population Health, University of Western Australia, Perth, WA, Australia
    Search for more papers by this author
  • JG Cecatti,

    1. Department of Obstetrics and Gynaecology, University of Campinas, Campinas, Brazil
    Search for more papers by this author
  • J Barrett,

    1. Sunnybrook Health Sciences Centre, Sunnybrook Research Institute, Toronto, ON, Canada
    Search for more papers by this author
  • K Jayaratne,

    1. Family Health Bureau, Ministry of Health, Colombo, Sri Lanka
    Search for more papers by this author
  • S Mittal,

    1. Department of Obstetrics and Gynaecology, Fortis Memorial Research Institute, Gurgaon, India
    Search for more papers by this author
  • E Ortiz-Panozo,

    1. Centre for Population Health Research, National Institute of Public Health, Cuernavaca, Mexico
    Search for more papers by this author
  • JP Souza,

    1. Department of Reproductive Health and Research, UNDP/UNFPA/UNICEF/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction (HRP), World Health Organization, Geneva, Switzerland
    Search for more papers by this author
  • C Crowther,

    1. Liggins Institute, The University of Auckland, Auckland, New Zealand
    Search for more papers by this author
  • E Ota,

    1. Department of Maternal and Child Health, National Centre for Child Health and Development, Tokyo, Japan
    Search for more papers by this author
  • R Mori,

    1. Department of Health Policy, National Center for Child Health and Development, Tokyo, Japan
    Search for more papers by this author
  • on behalf of the WHO Multicountry Survey on Maternal and Newborn Health Research Network


  • T Ganchimeg and N Morisaki are equal first authors.

Abstract

Objective

To describe the mode and timing of delivery of twin pregnancies at ≥34 weeks of gestation and their association with perinatal outcomes.

Design

Secondary analysis of a cross-sectional study.

Population

Twin deliveries at ≥34 weeks of gestation from 21 low- and middle-income countries participating in the WHO Multicountry Survey on Maternal and Newborn Health.

Methods

Descriptive analysis and effect estimates using multilevel logistic regression.

Main outcome measures

Stillbirth, perinatal mortality, and neonatal near miss (use of selected life saving interventions at birth).

Results

The average length of gestation at delivery was 37.6 weeks. Of all twin deliveries, 16.8 and 17.6% were delivered by caesarean section before and after the onset of labour, respectively. Prelabour caesarean delivery was associated with older maternal age, higher institutional capacity and wealth of the country. Compared with spontaneous vaginal delivery, lower risks of neonatal near miss (adjusted odds ratio, aOR, 0.63; 95% confidence interval, 95% CI, 0.44–0.94) were found among prelabour caesarean deliveries. A lower risk of early neonatal mortality (aOR 0.12; 95% CI 0.02–0.56) was also observed among prelabour caesarean deliveries with nonvertex presentation of the first twin. The week of gestation with the lowest rate of prospective fetal death varied by fetal presentation: 37 weeks for vertex–vertex; 39 weeks for vertex–nonvertex; and 38 weeks for a nonvertex first twin.

Conclusions

The prelabour caesarean delivery rate among twins varied largely between countries, probably as a result of overuse of caesarean delivery in wealthier countries and limited access to caesarean delivery in low-income countries. Prelabour delivery may be beneficial when the first twin is nonvertex. International guidelines for optimal twin delivery methods are needed.

Introduction

Twin gestations account for 0.5–2.0% of all pregnancies, with rates varying largely by race/ethnicity and country.[1] Twins are at a higher risk for preterm delivery,[2] and for infant morbidity and mortality,[3, 4] and many epidemiological studies and a recent multicentred randomised controlled trial (RCT) show that the optimal duration of pregnancy may be shorter in multiple pregnancies compared with singletons.[2, 5-8] Such outcomes are addressed in the current clinical guidelines of several countries.[9-12]

Recent studies state that there is no evidence of better perinatal outcomes for prelabour caesarean delivery versus planned vaginal delivery, as long as the first twin is in the vertex position in an otherwise uncomplicated pregnancy.[13-16] Despite the lack of evidence, there is a continuing increase in caesarean section rates for twins in many low-, middle- and high-income countries,[17-19] and as noted by Blickstein we are in danger of ‘a vicious circle in favour of caesarean section’,[13] through decreasing experience in vaginal delivery.[20]

On the other hand, sub-Saharan Africa seems to be excluded from this trend of increased rates of caesarean section,[21, 22] and in many low- and middle-income countries, inequalities in access to caesarean delivery still reflect socio-economic inequities for skilled delivery.

In this secondary analysis of the WHO Multicountry Survey on Maternal and Newborn Health—a large multicentre cross-sectional survey of deliveries in 29 countries in Africa, Latin America, Asia and the Middle East—we sought to describe the mode and timing of twin deliveries in institutions capable of caesarean delivery, and to address their association with perinatal outcomes.

Methods

Study design and data collection

This is a secondary data analysis of the WHO Multicountry Survey on Maternal and Newborn Health (WHOMCS) carried out in 359 health facilities from 29 countries in Africa, Asia, Latin America, and the Middle East. Full methodological details are available in our previous article.[23, 24] In brief, a multistage cluster sampling method was applied to acquire samples of health facilities in 29 countries. Trained health professionals at these facilities retrieved information from medical records, including individual data on demographics and reproductive characteristics, medical conditions during pregnancy, birth outcomes, complications, and received interventions for all women who were admitted for delivery or had severe maternal outcomes. In addition, health facility data were obtained, regarding capacity on laboratory tests, blood transfusions, availability of intensive care for mothers and babies, surgical procedures, human resources and training, and the capabilities of obstetrics and neonatal healthcare services. Data were collected in each facility over a period of 2–4 months from May 2010 to December 2011.

Study population

For our analysis we used a subpopulation of the available data set that included all twin pregnancies delivered at or later than 34 weeks of gestation in low- or middle-income countries, with no medical contraindication for vaginal delivery. As approximately 50% of twins are delivered preterm, with 30% delivered in the late preterm period (34–36 weeks of gestation),[9] a cut-off of 34 weeks of gestation was chosen in accordance with previous studies.[16, 25]

Of the 317,107 deliveries observed in this subpopulation, 4112 were twin deliveries. We excluded 468 twin deliveries in seven countries where data on all second twins were missing (Paraguay, Peru, Philippines, Qatar, Thailand, Vietnam, and Uganda). Thirty-eight deliveries in high-income countries (Japan) were included only in a descriptive analysis on variability between countries. We further excluded deliveries with congenital malformations or macerated fetal deaths of either twin (n = 82), maternal complications (n = 557), women with previous caesarean section (n = 389), as well as deliveries with missing data on gestational age, onset of labour, fetal presentation, birthweight, infant status at birth or at seventh day of life of either infant (n = 108), and retrieved 2134 twin deliveries at or later than 34 weeks of gestation that had no medical indication for urgent delivery.

As only 121 (5.6%) were deliveries after induction of labour, we also excluded these and focused on the remaining 2013 deliveries from 21 countries that were either prelabour caesarean delivery (n = 384) or delivery following spontaneous onset of labour (n = 1629) (Figure 1).

Figure 1.

Population flow chart.

Exposure and potential confounding variables

Our main comparison of interest was to compare prelabour caesarean delivery with ‘expectant management’: the decision to not perform a prelabour caesarean delivery at a specific gestational age (in which case the comparison groups would be all deliveries after that gestational age).[26, 27] We also defined intrapartum caesarean section as caesarean section after the spontaneous onset of labour.

The WHOMCS individual data set includes demographic characteristics, obstetric and medical history, mode of delivery, and maternal and perinatal outcomes until discharge from hospital, or up to the seventh postpartum day or death, whichever occurred first. Morbidity and mortality occurring post-discharge or during a subsequent readmission were not captured.

Medical conditions or complications, such as intubation, were recorded as binary variables (yes/no), and severity, time of onset, and management were not captured. All continuous variables, such as maternal age, were converted into categorical variables. Gestational age was recorded in completed weeks based on the best available obstetric estimate; the method of estimation was not recorded. Chorionicity and amnionicity were not available in the WHOMCS database, and thus were not accounted for in the analysis.

The maternal-level confounding factors included were: maternal age (<20, 20–35, >35 years), maternal education (0, 1–4, 5–9, ≥10 years), and parity (0, 1–2, ≥3). Perinatal-level confounding factors were fetal presentation (vertex–vertex, vertex–nonvertex, nonvertex first twin), infant concordance (concordant, discordant, both of severe small gestational age, SGA), and prematurity (preterm, term). Discordancy was defined as an inter-twin birthweight difference of over 20% of the smaller twin,[28-31] and severe SGA as a birth weight below the third percentile of the singleton reference, using a population-based standard developed by Mikoliaczyk and collegues,[32] which was applied to our data set.

Additionally, the analyses were adjusted for the Human Development Index (HDI), based on the 2012 rankings,[33] as well as for the facility capacity index category, a proxy for the capacity of the institution to provide essential obstetric care and additional services, which was calculated as the total score of services, and categorised into low, medium, and high.

Main outcomes and definitions

Neonatal adverse outcomes were the main outcome of interest: stillbirth, early neonatal mortality, and neonatal near miss. Neonatal near miss was defined as any of the following: any intubation (at birth or at any time within the first week of life); nasal continuous positive airway pressure (CPAP); surfactant administration; cardiopulmonary resuscitation (cardiac massage); any surgery; or the use of any vasoactive drug, anticonvulsants, phototherapy in the first 24 hours, steroids to treat refractory hypoglycaemia, or therapeutic intravenous antibiotics. This set of life-saving interventions has previously been used to study the applicability of the near-miss concept in newborn infants, developed by Avenant,[34] and explored as ‘management markers of severity’ by Pileggi-Castro et al.[35]

Analysis and statistical methods

First we described country-specific prelabour caesarean rates, intrapartum caesarean section rates, and the mean age of delivery for twin pregnancies, stratified by fetal presentation and HDI ranking. For this analysis we included deliveries in Japan and in countries where all second-infant data were missing.

Next we reported the frequencies for maternal and neonatal characteristics in prelabour caesarean deliveries and spontaneous deliveries. We also used multilevel multivariate logistic regression to calculate the association of each characteristic and the odds ratios of prelabour caesarean delivery. This method enabled us to take into account the clustering effect within mothers, facilities, and within HDI subgroups, as well as adjusting for maternal characteristics.

To calculate the effect of gestational age for delivery on neonatal adverse outcomes, we adopted the prospective risk approach originally proposed by Feldman,[36] and calculated the ‘prospective perinatal mortality’ according to Kramer.[37] Prospective fetal death was calculated as the probability of stillbirth in deliveries at or after a certain gestational age, and prospective perinatal mortality was calculated as the probability of either early neonatal death or stillbirth in deliveries at or after a certain gestational age. This approach allows us to compare the risk of adverse outcomes between different gestational lengths, as well as between the termination and the continuation of the pregnancy, and between prelabour caesarean section and expectant management, both at a given gestational age. Our small sample size limited us from adjusting for potential confounding factors in this analysis.

As perinatal mortality and the neonatal near-miss rate were lower in women undergoing prelabour caesarean delivery, compared with those receiving expectant care for most gestational ages between 34 and 40 weeks of gestation, we also compared prelabour caesarean section at any age above 34 weeks of gestation with women undergoing expectant management. We accounted for any clustering effects within mother, facility, and HDI subgroups, as well as adjusting for maternal characteristics other than gestational age; however, we did not adjust for gestational week in this analysis. Adjustment for gestational age would have created a comparison between prelabour delivery and spontaneous labour both at the same gestational age, when actually the correct comparison for prelabour delivery at a given gestational age is to directly compare prelabour delivery with women undergoing expectant management.[27, 38, 39] We repeated this with stratification by fetal presentation.

To make a direct comparison of prelabour delivery and expectant management, we compared prelabour delivery at 34–36 weeks of gestation with expectant management (i.e. delivery after 37 weeks of gestation), as well as prelabour delivery at 37–38 weeks of gestation with expectant management (delivery after 39 weeks of gestation). In this analysis we also used multilevel analyses to account for clustering within the mother, institutions, and HDI subgroups, and adjusted for maternal characteristics. We further investigated effect modification by HDI subgroups.

For all analyses, estimates were adjusted for survey design and P < 0.05 was considered significant. We have reported all odds ratios (ORs) with corresponding 95% confidence intervals (95% CIs). Missing values were excluded from all logistic regression models. Statistical analysis was conducted using stata/mp 13.0 (StataCorp LP, College Station, TX, USA).

Results

In Figure 2(A,B) we show the timing and mode of delivery in twin pregnancies for each country. Mean gestational age for delivery decreased and the proportion of prelabour caesarean deliveries increased with higher HDI: the coefficients of determination were R2 = 0.26 and 0.51, respectively.

Figure 2.

(A) Mean gestational age of twin delivery by country HDI. (B) Proportion of prelabour caesarean delivery by country HDI.

Figure 2(A) shows the average gestational length of delivery by the HDI ranking of the country. Of the 21 countries participating in our study, the average length of gestation for all twins delivered after 34 weeks of gestation was 37.6 weeks of gestation. In seven low- and middle-HDI countries (Afghanistan, Angola, Cambodia, Congo, Nicaragua, Niger, and Uganda), twin pregnancies have an average length of gestation above 38 weeks of gestation.

Figure 2(B) shows the prelabour caesarean section rate by the HDI ranking of the country, in which 16.8% of all twins were delivered by prelabour caesarean section, and another 17.6% were delivered through intrapartum caesarean delivery following the spontaneous onset of labour. The prelabour caesarean rate differed by HDI, with 11% in low-, 27% in medium-, 35% in high-, and 56% in very high-HDI countries, with rates in Japan, China, and Brazil of over 60%, and rates in Cambodia, Angola, Uganda, and Afghanistan of below 5%.

A more complete overview is shown in Table S1, where average gestational length and caesarean section rate (prelabour and emergency) are shown by fetal presentation for each country. The overall rate of prelabour delivery (19.1% overall) differed largely by presentation: 14% of deliveries were for vertex–vertex presentation; 17% were vertex–nonvertex presentation; and 32% occurred for nonvertex first-twin presentation. Rates of prelabour caesarean section for nonvertex first-twin presentation were significantly higher compared with vertex–vertex presentation in Ecuador, Jordan, Afghanistan, India, Pakistan, Kenya, and Nigeria, however.

Table 1 illustrates maternal, obstetric, and institutional characteristics, as well as their association with the estimated risks of receiving prelabour caesarean delivery. Of the pregnancies experiencing a spontaneous onset of labour, 25% were delivered through intrapartum caesarean section.

Table 1. Maternal, obstetric, and institutional characteristics and their associations with prelabour caesarean delivery. Analysis of 2013 twin deliveries at ≥34 weeks of gestation
 All deliveries n (%)Prelabour caesarean section n (%)Spontaneous labour n (%)ORaOR (95% CI)
  1. Adjusted odds ratio (aAOR), adjusted for maternal age, marital status, education, parity, fetal presentation, birthweight concordance, prematurity, facility capacity, and country HDI (as of 2012). Severe SGA: birthweight below third percentile of singleton reference.

  2. Level of significance: * < 0.05; ** < 0.01; *** < 0.001.

Number of deliveries 20133841629  
Age
<20 years112 (5.6)15 (3.9)97 (6.0)Ref.Ref.
20–34 years1587 (78.8)303 (78.9)1284 (78.8)1.522.58 (1.22–5.46)
≥35 years314 (15.6)66 (17.1)248 (25.2)1.723.62 (1.38–9.47)
Marital status
Single127 (6.4)23 (6.1)104 (6.4)Ref.Ref.
Married1872 (93.6)357 (93.9)1515 (93.6)1.071.40 (0.46–2.95)
Education
0 years504 (25.0)58 (15.1)446 (27.4)Ref.Ref.
1–6 years257 (12.8)32 (8.3)225 (13.8)1.090.786 (0.42–1.42)
7–9 years390 (19.4)65 (16.9)325 (20.0)1.540.79 (0.45–1.38)
10–12 years452 (22.4)87 (22.7)365 (22.4)1.841.07 (0.58–2.00)
>12 years410 (20.4)142 (37.0)268 (16.4)4.011.89 (1.07–3.34)
Parity
0624 (31.0)217 (56.5)407 (25.0)Ref.Ref.
1–2895 (44.5)111 (28.9)784 (48.2)0.260.30 (0.20–0.45)
≥3493 (24.5)56 (14.6)437 (26.8)0.240.25 (0.14–0.44)
Mode of delivery
Vaginal1223 (60.7)1223 (75.1)
Intrapartum caesarean406 (20.2)406 (24.9)
Prelabour caesarean384 (19.1)384 (100)
Fetal presentation
Vertex–vertex presentation969 (48.1)136 (35.4)833 (51.1)Ref.Ref.
Vertex–nonvertex presentation569 (28.3)96 (25.0)473 (29.1)1.241.24 (0.96–1.71)
Nonvertex first-twin presentation475 (23.6)152 (39.6)323 (19.8)2.893.85 (2.30–6.44)
Birthweight concordance
Concordant twin835 (41.5)179 (46.6)656 (40.3)Ref.Ref.
Discordant twin883 (43.8)169 (44.0)714 (43.8)0.870.90 (0.68–1.18)
Both severe SGA twins295 (14.7)36 (9.4)259 (15.9)0.510.52 (0.34–1. 13)
Prematurity
Preterm delivery (<37 weeks)1484 (73.7)411 (77.7)1218 (82.1)1.130.85 (0.60–1.21)
Term delivery (≥37 weeks)529 (26.3)118 (22.3)266 (17.9)Ref.Ref.
Facility capacity
High458 (26.4)139 (38.5)319 (23.2)Ref.Ref.
Medium708 (40.8)193 (53.5)515 (37.4)0.891.44 (0.72–2.88)
Low570 (32.8)29 (8.0)541 (39.4)0.120.37 (0.15–0.97)
HDI country groups
Very high and high492 (24.4)202 (52.6)290 (17.8)Ref.Ref.
Medium440 (21.9)52 (13.5)388 (23.8)0.190.17 (0.06–0.49)
Low1081 (53.7)130 (34.0)951 (58.4)0.200.23 (0.10–0.87)

For maternal characteristics, lower age and multiparity were associated with fewer prelabour caesarean deliveries. Mothers aged under 20 years and parous mothers were nearly three times more likely to experience the spontaneous onset of labour, compared with women aged over 20 years and nulliparous women.

Marital status was not significantly associated with the onset of labour, but those who had completed over 12 years of education were 1.9 times more likely to have a prelabour caesarean section.

Fetal presentation was highly associated with the onset of labour: a nonvertex first-twin presentation increased the odds of prelabour delivery by four times, compared with vertex–vertex presentation. The risk of intrapartum caesarean section after an attempted vaginal delivery was five times higher in a nonvertex first-twin presentation, compared with vertex–vertex twins.

Concordance was not significantly associated with the onset of labour, but women with small-for-gestational-age (SGA) twins were more likely to experience spontaneous labour.

The capacity of the healthcare facility and the level of development of the country were both associated with the onset of labour. Deliveries in low-capacity facilities and low-income countries were 2.5–5.0 times more likely to experience spontaneous labour.

Next we investigated the distribution of timing of twin deliveries, rates of prelabour caesarean section, as well as their perinatal outcomes for each gestational week by fetal presentation. The largest number of deliveries was at 37–38 weeks of gestation, which was the same when limited to only prelabour caesarean delivery. Most prelabour caesarean deliveries were performed at 38 weeks of gestation for vertex–vertex presentation, and at 37 weeks of gestation for vertex–nonvertex and nonvertex first-twin presentations. The lowest rate of prospective fetal death was observed at 36–37 weeks of gestation, and varied slightly by fetal presentation: 37 weeks of gestation for vertex–vertex; 39 weeks of gestation for vertex–nonvertex; and 38 weeks of gestation for nonvertex first twin. The prospective fetal death rate was higher in vertex–nonvertex and nonvertex first-twin presentations, compared with vertex–vertex twins, at every gestational age, except for 41 weeks of gestation. Detailed results are shown in Table 2.

Table 2. Distribution of fetal and neonatal adverse outcomesa by gestational age. Analysis of 2013 twin deliveries at ≥34 weeks of gestation
 Gestational age at birth by completed weeks [n (%)]
3435363738394041
  1. A neonatal near miss is considered as any of the following: any intubation (at birth or at any time within the first week); nasal CPAP; surfactant administration; cardiopulmonary resuscitation (cardiac massage); use of any vasoactive drug; use of therapeutic intravenous antibiotics; use of blood products; use of steroids to treat refractory hypoglycaemia; or any surgery.

  2. a

    Defined as events in one or both twins.

  3. b

    Calculated as (all fetal deaths occurring at or after given gestational age/all fetuses delivered at or after given gestational age) × 1000.

  4. c

    Calculated as (number of events at a given gestational age/total live births at a given gestational age) × 1000.

All deliveries, = 2013
Deliveries (n)12815524643344424031849
Prelabour caesarean section [n (%)]26 (20)36 (23)56 (23)85 (20)90 (20)40 (17)45 (14)6 (12)
Live births (n)24729447283385245960892
Prospective fetal deathb41.742.241.341.442.846.147.651.0
Early neonatal deathc32.427.231.840.843.441.439.554.3
Neonatal near missc161.9176.9178.0196.9186.6176.5190.8250.0
Vertex–vertex presentation, n  = 969
Deliveries (n)587411219923511515224
Prelabour caesarean section [n (%)]11 (19)13 (18)26 (23)20 (10)39 (17)13 (11)12 (8)2 (8)
Live births (n)11113821538745221729343
Prospective fetal deathb40.840.638.237.941.848.142.683.3
Early neonatal deathc34.52740.242.751.147.849.562.5
Neonatal near missc206.9243.2218.8226.1238.3221.7184.8333.3
Vertex–nonvertex presentation, = 569
Deliveries (n)394668126115689116
Prelabour caesarean section [n (%)]5 (13)9 (20)14 (21)24 (19)20 (17)12 (18)9 (10)3 (18)
Live births (n)768913123921713317431
Prospective fetal deathb87.990.693.096.293.180.0102.862.5
Early neonatal deathc39.522.515.341.832.337.640.232.3
Neonatal near missc184.2168.5167.9205.0170.5157.9224.1193.5
Nonvertex first-twin presentation, = 475
Deliveries (n)3135661089457759
Prelabour caesarean section [n (%)]10 (32)14 (40)16 (24)41 (38)31 (33)15 (26)24 (32)10 (11)
Live births (n)606712620718310914118
Prospective fetal deathb82.183.383.181.680.999.3107.10.0
Early neonatal deathc16.729.931.738.638.327.521.355.6
Neonatal near missc150.0164.2238.1227.1191.3183.5205.7277.8

As shown in Table S2, we also compared perinatal adverse outcomes between prelabour delivery and expectant management for each week of gestation from 34 to 41 weeks of gestation. Perinatal mortality was lower at each gestational week in prelabour caesarean deliveries compared with spontaneous labour deliveries. Our small sample size made comparison difficult after stratification by fetal presentation.

Following these analyses, we further compared perinatal adverse outcomes between prelabour caesarean delivery and spontaneous labour, adjusting for institutional, maternal, and neonatal characteristics, and stratifying by fetal presentation. Table 3 shows that the early neonatal near-miss rate and early neonatal mortality were 0.39 (95% CI 0.19–0.80) and 0.63 (95% CI 0.44–0.94) times lower, respectively, in twins delivered by prelabour caesarean delivery, compared with deliveries following the spontaneous onset of labour. As shown in Table 3, after stratification by fetal presentation this protective effect for reducing neonatal mortality was not present in vertex–vertex twins (1.07; 95% CI 0.44–2.57), and was stronger in nonvertex first-twin deliveries (0.12; 95% CI 0.02–0.56). All effects for neonatal near miss were insignificant after stratification, but produced similar estimates: 0.64 for vertex–vertex; 0.66 for vertex–nonvertex; and 0.59 for nonvertex first-twin presentation.

Table 3. (a) Risks of perinatal adverse outcomes associated with prelabour caesarean delivery, (b) by fetal presentation. Analysis of 2013 twin deliveries at ≥34 weeks of gestation
(a)
 Total [n (%)]Prelabour caesarean delivery [n (%)]Spontaneous labour [n (%)]ORaOR (95% CI)
Fetal death168 (8.4)30 (7.8)138 (8.5)0.920.92 (0.56–1.50)
Early neonatal death150 (7.5)15 (3.9)135 (8.3)0.730.39 (0.19–0.80)
Neonatal near miss647 (32.1)103 (26.8)544 (33.4)0.450.63 (0.44–0.94)
(b)
 Fetal deathEarly neonatal deathNeonatal near miss
Total [n (%)]Prelabour caesarean section [n (%)]aOR (95% CI)Total [n (%)]Prelabour caesarean section [n (%)]aOR (95% CI)Total [n (%)]Prelabour caesarean section [n (%)]aOR (95% CI)
  1. Adverse outcomes defined as events in one or both twins. Neonatal near miss is considered as any of the following: any intubation (at birth or at any time within the first week); nasal CPAP; surfactant administration; cardiopulmonary resuscitation (cardiac massage); use of any vasoactive drug; use of therapeutic intravenous antibiotics; use of blood products; use of steroids to treat refractory hypoglycaemia; or any surgery. Adjusted odds ratio (aOR), adjusted for: maternal age; marital status; education; parity; birthweight concordance; prematurity; facility capacity; and country HDI.

  2. Levels of significance: * < 0.05; ** < 0.01; *** < 0.001.

All deliveries168 (100)30 (100)0.92 (0.56–1.50)150 (100)15 (32.1)0.39 (0.19–0.80)647 (100)103 (100)0.63 (0.44–0.94)
Vertex–vertex presentation, n = 96979 (47.0)8 (26.7)0.61 (0.23–1.62)86 (57.3)10 (66.7)1.07 (0.44–2.57)300 (46.4)37 (35.9)0.64 (0.37–1.10)
Vertex–nonvertex presentation, n = 56950 (29.8)8 (26.7)0.73 (0.26–2.05)36 (24.0)2 (13.3)0.12 (0.01–1.06)183 (28.3)28 (27.2)0.66 (0.34–1.26)
Nonvertex first-twin presentation, n = 47539 (23.2)14 (46.8)1.09 (0.48–2.45)28 (18.7)3 (20.0)0.12 (0.02–0.56)164 (25.3)38 (36.9)0.59 (0.30–1.15)

Table 4 shows a direct comparison between prelabour caesarean section and expectant management, stratified by the HDI of the country. The perinatal mortality and neonatal near-miss rates were lower at 37–38 weeks of gestation for prelabour caesarean delivery, and higher for expectant management, when compared with 34–36 weeks of gestation. Although all estimates were nonsignificant, perinatal mortality and the neonatal near-miss rate were estimated to be 1.2–1.9 times higher at 37–38 weeks of gestation in the expectant management group, compared with prelabour caesarean delivery, for all HDI subgroups. When comparing perinatal mortality and the neonatal near-miss rate between prelabour caesarean delivery and expectant management at 34–36 weeks of gestation, estimates ranged from 0.8 for medium-HDI countries to 1.1 for very high- and high-HDI countries.

Table 4. Risk of perinatal mortality and neonatal near miss in expectant management, compared with prelabour caesarean delivery at given periods (34–36 versus 37–38 weeks of gestation). Analysis of 2013 twin deliveries at ≥34 weeks of gestation
 34–36 weeks37–38 weeks
  1. a

    Denominator is all fetuses delivered after given gestational age, or were delivered at given gestational age following spontaneous onset of labour.

  2. Level of significance: * P < 0.05; ** P < 0.01; *** P < 0.001.

Prelabour caesarean delivery
Number of deliveries (n)118175
Perinatal mortality and neonatal near miss [n (%)]49 (41.5%)58 (33.1%)
Expectant management a
Number of deliveries (n)1484607
Perinatal mortality and neonatal near miss [n (%)]589 (39.7%)247 (40.7%)
Odds ratio comparing expectant management with prelabour caesarean delivery, for each period. Stratified by Human Development Index (HDI), as of 2012 (95% CI) a
All countries0.92 (0.63–1.36)1.38 (0.97–1.97)
Very high- and high-HDI countries1.12 (0.62–2.02)1.23 (0.63–2.41)
Medium-HDI countries0.79 (0.39–1.61)1.89 (1.00–3.59)
Low-HDI countries0.99 (0.45–2.19)1.64 (0.89–3.01)

Discussion

Main findings

We observed a higher proportion of prelabour caesarean delivery in twin pregnancies of older women, women with a higher level of education, women who delivered at higher capacity facilities, and women from wealthier countries. Overall, the neonatal near-miss rate was lower among prelabour caesarean deliveries compared with expectant management. After adjusting for institutional, maternal, and neonatal characteristics, prelabour caesarean delivery was associated with lower neonatal mortality, perinatal mortality, and neonatal near-miss rate for nonvertex first-twin presentation. We also found the wealth of a country to be associated with higher rates of prelabour caesarean section and a lower mean gestational age for delivery in twins.

Strengths and limitations

To the best of our knowledge, this is the first multicountry study to describe the timing and mode of delivery for twin pregnancies in low- and middle-income settings. To take account of heterogeneity in different settings, we stratified countries by HDI and adjusted for health institutional capacity. To correctly calculate the risk of adverse outcomes at different gestational lengths, as well as to compare between termination and continuation of the pregnancy, we followed the prospective risk approach originally proposed by Feldman,[36] which has been recommended in other work.[7, 37]

Our study has two main limitations, however. First, as our study was conducted in relatively large health institutions capable of performing caesarean section, our findings may not be generalisable to deliveries in smaller facilities, especially in those with limited capacity of performing surgery or other more complex interventions, and our calculated proportions and averages may differ from those at the national level. As multiple pregnancies are considered to be high risk, however, and are often referred to higher institutions, we believe our study is likely to reflect existing practice and the risk of adverse perinatal outcomes. Second, as a result of a relatively small number of deliveries and events, our study lacks the scope to detect significant differences, and our estimations are unstable when stratified by each gestational week.

Third, we were not able to incorporate data on chorionicity or amnionicity, as these data were not collected. Although these are inevitably important factors when managing twin pregnancies, in many developing countries this information is not available before delivery because of the low consultation and ultrasound rate during the first trimester.

Interpretation

We found that countries with a lower HDI had lower rates of prelabour caesarean section and higher mean gestational age at delivery. Seven of the 21 low- and middle-income countries in our study had a mean length of gestation above 38 weeks of gestation. Although we had restricted our population to deliveries later than 34 weeks of gestation, this is longer than previously reported in high-income countries,[2, 5, 7, 40, 41] and is also higher in lower HDI settings.

Although longer gestational length may provide time for maturation of twin fetuses, the risk of perinatal death may outweigh this benefit beyond 37–38 weeks of gestation. Our study showed that expectant management after 37–38 weeks of gestation could increase perinatal morbidity and mortality. Perinatal outcomes may be improved in such cases if twins are closely monitored and if delivery is performed by 37–38 weeks of gestation. This is supported by studies in developed countries that balanced the risk of intrauterine stillbirth and neonatal morbidity,[2, 5-8] as well as a recent multicentred RCT that found planned twin deliveries at 37 weeks of gestation did not increase adverse outcomes,[8, 42, 43] and is reinforced by current guidelines that recommend twin delivery at 34–38 weeks of gestation.[9-11]

In most current guidelines on the management of twin pregnancies, fetal presentation is the determining factor for mode of delivery, with maternal characteristics such as maternal age, multiparity, and natural conception modifying the beneficial effect of prelabour caesarean delivery.[44] Several countries have specific recommendations for vertex–vertex or nonvertex first-twin presentations.[3, 11, 12] An attempted trial of vaginal delivery is recommended for vertex–vertex presentation, whereas a prelabour caesarean section is recommended for nonvertex first-twin presentation.[10, 12, 13, 45] In the case of vertex–nonvertex presentation, there is little available evidence upon which to base recommendations for delivery methods.[20, 46] A recent large trial of over 1400 women with twins showed that planned vaginal delivery was equally as feasible as planned caesarean section when delivering at 38 weeks of gestation;[47] however, the recommended method of delivery is still under debate. Observational studies using existing data sets covering births of routine clinical practice should be further explored for supportive evidence.

Our findings, as supported by other studies,[17, 20-22] show that experience and preference, as well as disparities in access to care, may be significant determinants as to who delivers by prelabour caesarean delivery. Only seven (25%) countries had significantly higher rates of prelabour caesarean section for nonvertex first-twin presentation, compared with vertex–vertex presentation. On the other hand, the proportion of prelabour caesarean deliveries was higher in older women, with a higher level of education, and in those admitted to higher capacity facilities and in higher HDI countries.

The difference by HDI was prominent. In five countries, prelabour caesarean section was performed in over 50% of vertex–vertex presentations, whereas in 11 countries the procedure was performed in less than 10% of such cases.

Surprisingly, differences were also observed in deliveries that featured a nonvertex first-twin presentation, a state where prelabour caesarean delivery is recommended in multiple guidelines,[12-14] because of the possibility of the fatal complication of interlocking twins and the high rate of intrapartum caesarean section after attempted vaginal delivery. In low-HDI countries 75% of women with a nonvertex first-twin presentation experienced the spontaneous onset of labour, 40% of whom required an intrapartum caesarean section.

A recent study in France on mothers undergoing a trial of vaginal delivery also reported that nearly half of such mothers required intrapartum caesarean section,[45] and made recommendations to avoid attempting vaginal delivery for nonvertex first-twin presentation. Our analysis also showed that prelabour caesarean section was associated with reduced neonatal morbidity and mortality in non-vertex twins. Although experience and high multiparity may contribute to the high success rate of vaginal delivery in 60% of low- and middle-income countries (in high-HDI countries, the success rate of vaginal delivery was 5%), an increase in prelabour caesarean section for nonvertex twin pregnancies would most likely lead to improved perinatal outcomes.

Many reports state that the human resources required for caesarean deliveries are still lacking in many low- and middle-income countries,[21, 48] which is one of the likely reasons why vaginal deliveries are attempted for cases where such delivery is not normally recommended: for example, in the case of nonvertex first-twin presentation. Our previous study shows that twin pregnancy is a significant risk factor for maternal and perinatal morbidity in low- and middle-income countries.[49] In light of this high risk and the 3–5% prevalence of multiple births, as well as the protective effect of prelabour caesarean delivery on the fetus, it is critical to develop measures that advance timely access to facilities capable of performing safe caesarean delivery in order to greatly improve twin pregnancy outcomes.

Conclusion

We found that the prelabour caesarean delivery rate was lower in countries with a low-HDI ranking, and that prelabour caesarean delivery may be beneficial for twins with nonvertex first-twin presentation.

Further studies, as well as the development of international guidelines on the optimal delivery method of twins, such as those implemented in developed countries, combined with training for the antepartum diagnosis of fetal presentation, could play an important role in enhancing maternal and neonatal outcomes in the management of twin pregnancy.

Disclosure of interests

None.

Contribution to authorship

NM and TG initiated the idea and designed the analysis plan, to which JB, JGC, JPV, KJ, and EO provided substantial advice. TG and NM performed the data analysis. NM and TG interpreted the results, wrote the article, and made revisions. Additionally, CC, JEOP, SM, JPS, and RM contributed to revisions, and approved the final version of the article.

Details of ethics approval

The HRP Specialist Panel on Epidemiological Research reviewed and approved the study protocol for technical content. This study was approved by the WHO Ethical Review Committee and the relevant ethical clearance mechanisms in all countries (protocol ID, A65661; date of approval, 27 October 2009).

Funding

This study was financially supported by the UNDP/UNFPA/UNICEF/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction (HRP); World Health Organization (WHO); United States Agency for International Development (USAID); Ministry of Health, Labour and Welfare of Japan; and Gynuity Health Projects. The sponsors had no role in the data collection, analysis, or interpretation of the data, the writing of the report, or the decision to submit for publication. All authors had access to the analysis plan, the outputs of that analysis, and could see the data if they wished to do so. All authors participated in the final discussion and approved the submission.

Acknowledgements

We would like to thank Emma Barber of the National Centre for Child Health and Development for her assistance in revising and editing the draft. The Multicountry Survey on Maternal and Newborn Health is a research project implemented by the WHO in a global network of health facilities between 2010 and 2011. This project is part of the WHO response to the United Nations Secretary General's call for action to improve Women's and Children's health around the world. In this regard, the WHO is grateful to the extensive network of institutions and individuals who contributed to the project design and implementation, including researchers, study coordinators, data collectors, data clerks, and other partners, including the staff from the Ministries of Health and WHO offices.

Ancillary