Pregnancy outcomes in preterm multiple gestations: Results from a prospective study in India and Pakistan (PURPOSe)

To evaluate perinatal outcomes in preterm multiple compared with singleton pregnancies in India and Pakistan.


| I N TRODUC T ION
Multiple gestation accounts for between 0.5% and 2.0% of all pregnancies, with the vast majority of multiple gestations being twins.][3][4][5][6][7][8][9][10] Rates of neonatal mortality can be up to six times higher in infants from multiple pregnancies compared with infants from singleton pregnancies, largely due to the early gestational age (GA) of the infants and small for GA (SGA), especially for the second twin.][3][4][5][6][7][8][9][10] Many epidemiological studies and a recent multicentre trial suggest that the optimal duration of pregnancy may be shorter for infants of multiple pregnancies than for those from singleton pregnancies.
][3][4][5][6][7][8][9][10] The global incidence of multiple birth has increased more than 70% in the last three decades, mostly due to the increased use of assisted reproductive technologies, but increased pregnancies in older women may be partially responsible.
The present study (PURPOSe) was initiated to understand the causes of death in stillbirths and preterm neonates in south Asia. 11Different from many earlier studies, we also used minimally invasive tissue sampling (MITS), 12,13 histological evaluation of internal organ tissues and the placenta, 14 as well as multiplex pathogen polymerase chain reaction (PCR) 15 to provide expert panels in India and Pakistan with sufficient data to evaluate the causes of death. 16,17

| M ET HODS
PURPOSe was a prospective, observational study using standardised data collection methods conducted in two sites in south Asia.The sites were located in Davangere, India (three public and private hospitals) and in Karachi, Pakistan (one public maternity and one children's hospital).These sites were selected to represent two south Asian areas, based on the relatively high rate of preterm birth in the region and the availability of experienced research teams.The study methods have been described in detail elsewhere. 11The primary goal of the overall study was to determine the cause of stillbirths and neonatal death in preterm neonates.

| Study population
At participating study hospitals, research staff screened all pregnant women at the time of presentation for delivery.Pregnant women with an expected or known preterm birth, defined as <37 weeks GA, and mothers of all stillbirths were approached for enrolment at the time of admission to the hospital.Women ≤14 years of age and those who did not provide consent were excluded.

| Data collection
At the time of delivery, study staff collected results of the physical examination, medical and obstetric history, as well as the clinical status and procedures conducted during the current delivery.GA was based on clinical ultrasound or, when unavailable, last menstrual period (LMP) and by clinical assessment, using a hierarchy based on the American College of Obstetrics and Gynecology guidelines. 18The placenta and cord blood were collected for all enrolled deliveries using standardised operating procedures (SOPs).Trained pathologists conducted a gross evaluation and histopathology evaluation of the placentas using the Amsterdam Consensus criteria. 14ewborns were weighed and their length measured after delivery by trained staff.In addition, for neonatal deaths, a physical examination was conducted at the time of death.With additional consent, trained study staff conducted an MITS, 12,13 which included a needle biopsy for lung, liver and brain tissues, as well as blood and cerebrospinal fluid (CSF), collected using standardised procedures.MITS samples were assessed using histological and molecular microbiological methods. 12,13The TaqMan Array Card (TAC) polymerase chain reaction (PCR) testing, which had been developed and validated by the U.S. Centers for Disease Control (CDC) for 75 pathogens, was used to evaluate organisms in the fetal tissues, cord, membrane and placental tissues. 15At the Indian site, with additional consent, a standardised complete autopsy was also performed.For infants who died after discharge, a verbal autopsy was performed.

| Cause of death (COD) determination
Preterm neonatal deaths with a birthweight ≥1000 g and stillbirths with a birthweight ≥1000 g were evaluated by independent panels to determine the cause(s) of death.The panellists determined the primary maternal, placental and neonatal COD as well as the contributing causes using standard procedures described in detail elsewhere. 16,17or each case, a pair of panellists, who were not directly involved with the conduct of the study, reviewed the cases.A brief clinical description of the case as well as all positive clinical maternal, fetal and placental findings, and results of the PCR bacteriological investigation and MITS histology were included in each report. 16,17Additionally, reference weight measures including the 10th percentile for GA based on the INTERGROWTH-21st criteria and mean placental weight for GA were presented in a standardised data report. 19ach pair of panellists reviewed their cases, with any discrepancies resolved through discussion among the team.Causes of death used nomenclature from the ICD-10 PM system. 20sing this system, infectious causes of death were identified as congenital infections, which were acquired prior to delivery, and neonatal infections, determined to have been acquired following delivery.

| Statistical analyses
The planned study sample size was 350 preterm neonatal deaths in both India and Pakistan to provide sufficient precision to detect causes of death that contributed to 20% or more of the deaths with 95% power.To reach the target in neonatal deaths, assuming the reported mortality rates, we attempted to screen and enrol all women admitted to study hospitals in preterm labour or for a preterm delivery over an 18-month period.All data were entered and reviewed at each site.Descriptive analyses were performed using SAS (SAS v.9.4).

| R E SU LTS
The population in this study included mothers of preterm pregnancies and their fetuses/infants from India and Pakistan who consented to be evaluated in the PURPOSe study.For this study, the mothers had to deliver a preterm infant, whether singleton or multiple, regardless of the number of infants and whether the infants were stillborn or live born or lived or died in the neonatal period.All term births were excluded.
Over the time-period of the project, there were 3912 preterm pregnancies where the women consented to participate and were eligible for this study.These mothers gave birth to 3615 (92.4%) singleton babies, 281 (7.2%) sets of twins, 14 (0.4%) sets of triplets and one set of quadruplets.
After exclusion of multiple births greater than two and 15 sets of twins with only one infant enrolled, our study population included 534 twin infants and 3615 singleton infants (Table 1).Of the singleton infants, 691 (19.1%) were stillborn and 2924 (80.9%) were live born.Of the 534 infants from twin pregnancies, 41 (7.7%) were stillborn and 493 (92.3%) were live born.Of the 267 sets of twins, in 14 cases (5.2%) both were stillborn, in 13 cases (4.8%) one was stillborn and one live born, and in 240 cases (90.0%) both were live born.Thus, for the analyses of live borns, there were 2924 liveborn singleton and 493 live-born twin infants in our study population.
Table 2 shows the characteristics of mothers of preterm twin pregnancies compared with mothers of preterm singleton pregnancies.The mothers of twins (n = 267) and the mothers of singletons (n = 3615) had similar age distributions, education level and gravidity.Hypertensive disorders and antepartum haemorrhage were more common in preterm singleton deliveries than in twin deliveries.
The twin infant deaths had a mean birthweight of 1170 g compared with 1442 g for the preterm singleton deaths (P < 0.0001).Of the preterm twin infants, 55.0% were female compared with 46.7% of the preterm singleton infants (P = 0.0021).
We also evaluated the stillbirths from the preterm singleton and twin pregnancies.Of the 3615 singletons, there were 691 (19.1%) stillbirths, and of the 534 preterm twin infants, there were 41 (7.7%) stillbirths.
We also compared neonatal mortality rates in live-born twins vs singletons by sex, gestational age, birthweight and SGA status (Table 4).Male preterm twin infants had a slightly higher mortality rate (32.0%) than male preterm singleton infants (23.4%), but the difference was not significant, and the mortality was rate was similar for female preterm twin infants (22.9%) vs preterm singleton infants (20.1%).A few (6.3%) singleton infants 20-27 weeks survived but no twin infants in that GA group survived.Singleton infants 28-31 weeks appeared more likely to die (55.6%) than twin infants (47.7%) in the same GA range, but the difference was not significant.There were no significant differences in mortality between twin and singleton infants by birthweight.Neonatal mortality was similar in both groups for SGA infants, but was higher (but not significantly so) for non-SGA twin infants (27.6%) compared with non-SGA singleton infants (20.9%).
Among the preterm live-born infants with a cause of death determination by the panel (Table 5) in both preterm twins and preterm singletons, the three most common causes of death were intrauterine hypoxia, infections acquired prior to birth and infections acquired at or after birth.The preterm twin infants appeared less likely to have died from intrauterine hypoxia but were more likely to have died from infections acquired at or after birth.Respiratory distress syndrome (RDS) was rarely considered by the panel to be the primary cause of death in either the twins (9.6%) or singletons (9.7%).Congenital anomalies were also rarely judged to be the cause of death in either the preterm twins (n = 2, 2.4%) or singletons (n = 27, 5.3%); the differences were not significant.a Relative risks (RR) of neonatal mortality between twins and singletons along with associated 95% confidence intervals and P-values for each level of a characteristic are from generalised linear models with site, twin status, characteristic and the interaction of twin status and characteristics.
T A B L E 5 Cause of death of preterm live-born ≥1000 g twin and singleton neonates.

Preterm singleton neonatal deaths P-value a
Neonatal deaths with cause of death b 178 neonatal deaths ineligible for a panel cause of death due to birthweight <1000 g or missing.c Other causes/conditions include intraventricular haemorrhage (IVH) of the newborn, intracranial haemorrhage of the newborn -unspecified, meconium aspiration, pulmonary haemorrhage and other ill-defined cause.d Multiple neonatal conditions that contribute to death may be selected.
When contributory causes of death were evaluated, intrauterine hypoxia, RDS and congenital infections were the most common contributory causes; other causes of death such as IVH, meconium aspiration syndrome or pulmonary haemorrhage were less common.These conditions generally occurred in similar proportions in preterm twin and preterm singleton neonatal deaths.

| DISCUS SION
The study population included 3882 mothers and their 4149 preterm fetuses/infants.Of the twins, 89.9% of mothers had two live births, 5.2% had two stillbirths and 4.9% had one live-born and one stillbirth.Thus, of the twin infants/fetuses, 7% were stillborn and 93% were born alive.The most common causes of neonatal death were birth asphyxia and infections, with the infection-related deaths about equally divided between those apparently acquired before or at/after birth.Results were reasonably consistent between India and Pakistan.That birth asphyxia was more common than RDS as a cause of death in this population focusing on twins is consistent with our prior findings. 21,22he much higher mortality in live-born preterm twins than in live-born singletons may be due to an increased risk of death for twins of similar gestation age or birthweight or may be due to the fact that twins are far more likely to be born preterm or have a low birthweight.In this study, although there were small differences in mortality between twins and singletons of similar birthweight and GA, that twins were far more likely to be born preterm explains most of the difference in preterm neonatal mortality between singleton and twin preterm live births.
Prematurity and low birthweight are more frequent in twin babies.Earlier studies have shown this risk was eight times higher in twins than in singletons.[3][4][5][6][7][8][9][10] In one study, the mortality among twin babies was 2.5 times higher than singletons for first-born twins and 3.5 times higher than singletons for second-born twins. 10In the current study, we observed a higher risk for fetal death in preterm singletons than in preterm twins, but a lower risk of fetal death in singletons.In a recent study, [7][8][9][10] the overall perinatal mortality rate was higher in twin pregnancies than in singleton pregnancies, which was most likely caused by the high preterm birth rate in twins.Leading causes of death in twins included infection (29.5%), intrapartum hypoxia (25.6%) and RDS (8.5%).Causes of deaths in our study in twin babies were similar to those in preterm singleton neonatal deaths.We also noted that the percentage of deaths in live-born twins and singletons for those diagnosed with RDS and asphyxia was similar.Twins had a smaller percentage of deaths due to anomalies than singletons did.When infections as a cause of death were explored, preterm twin infants were more likely to die due to infections acquired after birth than were preterm singletons, probably because of a prolonged intensive care unit stay.
This study had its strengths and weaknesses.The strengths include the large number of multiple births evaluated, that the data were derived from a clearly defined population and that a comparison group of preterm singletons was available and also clearly defined.Potential weaknesses include that all cases were recruited at delivery so that potentially useful data normally available during the prenatal period may not have been available.The gestational ages for both the singleton and twin pregnancies were based on all available data but often did not include ultrasound estimations of gestational ages, so that calculations involving gestation age such as the preterm birth rate and rates of SGA were more likely to be inaccurate than if the ultrasound examinations had been performed.In addition, we were unable to distinguish monochorionic from dichorionic twin pregnancies.This is relevant due to the significant differences in outcomes related to preterm birth and stillbirth.The underlying mechanisms between these two types of pregnancies, such as twin-to-twin transfusion syndrome, selective intrauterine growth restriction and twin reversed arterial perfusion sequence (TRAP sequence), can lead to preterm birth and stillbirth by different mechanisms.

| CONCLUSIONS
In the PURPOSe Study, infants from preterm twin pregnancies were associated with significantly higher rates of adverse neonatal outcomes than infants from preterm singleton pregnancies.However, neonatal mortality rates in preterm twins than in singletons when evaluated by infant sex, GA, birthweight and SGA, were generally similar to rates of preterm singleton neonatal mortality.Thus, the higher rate of neonatal mortality in live-born twin infants is related to the fact that these infants were more likely to be born preterm rather than to any inherent characteristics of the babies themselves.The higher mortality in live-born twin pregnancies suggests that identification of these pregnancies and referral to facilities with capacity for quality obstetric and newborn care should be considered.

AU T HOR C ON T R I BU T ION S
GG, EMM and RLG conceived the analyses and wrote the initial draft.SST, SS, SSG, SMD, GG, HY and SY oversaw study implementation.Protocol was developed by SS, SSG, GG, RLG, AA, RMS, EMM and JK.KH and EMM conducted the analyses.All authors reviewed and approved the final article.

AC K NO W L E D GE M E N T S
The study was led by the PURPOSe study Group (Appendix).We thank the leadership of the clinical sites and the subjects at each site who participated in this study.

DATA AVA I L A BI L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

E T H IC S A PPROVA L S
The study was reviewed and approved by the ethics review committees at participating sites: the Aga Khan University (Karachi, Pakistan), KLE Academy of Higher Education and Research (Belagavi, India), J.J.M. Medical College (Davangere, India) and RTI International (Durham, NC, USA).All women provided informed written consent for themselves and their neonate prior to participation in the study.

PAT I E N T C ONSE N T
Women provided written informed consent prior to participation in the study.

T A B L E 1
The study population for the multiple pregnancy study.Maternal Preterm deliveries (n = 3897) Quadruplets n = 1 Triplets n = 14 Twins n = 267 Singletons n = 3615 Of the 267 twin deliveries Stillbirth and live birth = 13 Both stillbirths = 14 Both live births = 240 Fetuses and infants Of the 4149 fetuses and infants Singleton infants (n = 3615) Stillbirths = 691 Live births = 2924 Twin infants (n = 534) Stillbirths = 41 Live births = 493 T A B L E 2 Maternal characteristics of preterm live-born twins (defined as at least one live born) compared with mothers of singletons.

F
U N DI NG I N FOR M ION This study was funded through grants from the Bill & Melinda Gates Foundation.

Variable Preterm twin neonatal mortality, n/N (%) Preterm singleton neonatal mortality, n/N (%) RR a P-value a
Characteristics of preterm live-born twins compared with singletons.
T A B L E 3 a Hypothesis test results: P-values calculated from regression models adjusted by site for birthweight and gestational age, and Cochran-Mantel-Haenszel test adjusted by site (General statistic) for all other variables.b Excludes 13 stillbirths from twin stillbirth/live-birth dyads.T A L E 4 Neonatal mortality in live-born twins compared with singletons by sex, GA, birthweight and SGA.