Twin Births In The Pelotas (Brazil) 2004 And 2015 Birth Cohorts 


 Background: In the last decade, there has been an increase in the number of multiple pregnancies. Our aim was to describe the prevalence and duration of multiple pregnancies and compare first- and second-born twins to each other and to singletons, in terms of unfavorable perinatal outcomes and infant mortality rate (IMR). Methods: The 2004 and 2015 Pelotas Birth Cohorts are population-based studies that enrolled all newborns of mothers living in the urban area of Pelotas, South Brazil. All five maternity hospitals in Pelotas were visited daily by the research team, from January 1st to December 31st in 2004 and 2015. A monitoring system was assembled to detect all deaths of cohort participants in the years 2004, 2005, 2015, and 2016. Twinning rate was described according to maternal characteristics (family income, schooling, age, skin color, and morbidity) and parity. First- and second-born twins were compared to each other and to singletons, according to low birthweight (LBW), 5-minute Apgar <7, admission to neonatal intensive care unit (NICU), and IMR:1000 live births (LB). Results: Among 4,187 pregnancies in 2004 and 4,220 in 2015, respectively, 42 (1.0%) and 56 (1.3%) were multiple. Eighty-four twins were born alive in 2004 and 111 in 2015. In the two cohorts, no maternal characteristics were associated with the occurrence of multiple pregnancies. Twin births <34 weeks gestational age more than doubled from 2004 (19.0%) to 2015 (42.1%) (p=0.03). In both cohorts, LBW and admission to the NICU were more frequent in twins than in singletons, with no difference between first- and second-born twins. There was no difference between first-born and second-born twins at the two cohorts nor between twins and single-born infant mortality at the 2004 cohort. Among second-born twins in 2015, the post-neonatal mortality rate was 8 times higher than in singletons (37.7:1,000 LB versus 4.8:1,000 LB), and infant mortality rate (IMR) was 6 times higher (75.4:1,000 LB versus 12.5:1,000 LB). Conclusion: More than one out of 50 births in Pelotas was a twin, generally born prematurely and with a risk six times higher than singletons of dying alongside the first year of life.


Introduction
Twin births represent 2-4% of pregnancies in the world. (1) In recent decades, due to the growing popularity of assisted reproduction techniques such as in vitro fertilization and intrauterine insemination, there has been an increase in the number of multiple pregnancies, especially in developed countries. (2) In the United States, for example, the twinning rate increased 78% in the last 30 years. (3) In Brazil, according to data from the Information System on Live births (SINASC), the national rate is 1.13%, with regional variations, reaching 2.5% in some of Brazil's larger cities. (4) Twin pregnancies are associated with higher risks of preterm birth, low birthweight (LBW), intrauterine growth restriction, and umbilical cord prolapse.(5, 6) As a result, the intrapartum and perinatal mortality rates are three and seven times higher, respectively, among twins compared to singletons. (7) Likewise, maternal mortality and other serious conditions such as eclampsia, premature rupture of membranes, and placental abruption are three times more frequent in twin pregnancies. (8)(9)(10) Birth order in twin pregnancies has also been investigated as a risk factor for newborns, and lower birthweight and higher perinatal mortality were observed in second-born compared to rst-born twins, (6,(11)(12)(13)(14) mainly due to hypoxia and prolonged labor. (15)(16)(17)(18) There are still few studies on the epidemiology of twin pregnancies and twin births in Brazil. (4,8,(19)(20)(21)(22) Thus, based on data from the 2004 and 2015 Pelotas Birth Cohorts, this study aimed to describe the prevalence of multiple pregnancies according to maternal characteristics and to compare rst-and second-born twins from the same cohort to each other and to newborns from singleton pregnancies, according to the incidence of unfavorable perinatal outcomes and infant mortality rate (IMR). The use of two populationbased birth cohorts from the same city, eleven years apart, with data collected for each twin separately at the time of birth, allows to estimate change in twin births prevalence and to assess difference of risk according to birth order.

Methods
Two cohort studies in Pelotas, Brazil, aimed to include all hospital births in the city in the years 2004 and 2015. The recruitment strategies and data collection methods were very similar in the two cohorts. (23,24) Brie y, from January 1st to December 31st in 2004 and 2015, all ve hospitals with maternity departments in the city of Pelotas were visited daily by the research team. All newborns of mothers living in the urban area of the municipality were eligible to the cohorts. Children born outside the hospital (fewer than 2% in the two cohorts) were also included, since the mothers normally came to the maternity department soon after giving birth, when they were included in the study.
The mothers were interviewed during their hospital stay (perinatal survey) in the rst 24 hours postpartum by trained interviewers, who applied previously tested structured questionnaires. Enrollment totaled 4,231 children in 2004 and 4,275 in 2015.
Newborns were weighed by the hospital team using pediatric scales, SECA model 376, which were calibrated regularly by the study team. LBW was de ned as birthweight less than 2,500 grams and extremely low birthweight (ELBW) as birthweight less than 1,500 grams. Length was measured by the research team using an ARTHAG infantometer in 2004 and a Harpenden infantometer in 2015, both with precision to 1 mm. In the 2004 cohort, gestational age was calculated from the date of last menstruation (DLM), as long as consistent with the infant's weight, length, and head circumference at birth, in relation to the normal curves for these parameters for each week of gestational age. (25) If the gestational age based on DLM was unknown or inconsistent, maturity was estimated by ultrasound, as long as it had been performed prior to the 20th gestational week and recorded on the pregnant woman's prenatal card, and based on the Dubowitz method, (26) which was applied to all the newborns. In the 2015 cohort, gestational age was calculated on the basis of DLM and obstetric ultrasound tests (about 85% of the mothers had ultrasound records from the rst or second trimester). (27) For the analysis, gestational age was divided into three categories: < 34, 34-36, and ≥ 37 weeks. Five-minute Apgar score less than 7 and admission to neonatal intensive care unit (NICU) were extracted from the hospital les. Birth order of twins was obtained from the time of birth recorded on the newborn les.
A monitoring system was assembled to detect all deaths of cohort participants in the years 2004, 2005, 2015, and 2016. Regular visits were made to the city's hospitals, where intensive care units, maternity departments, pediatric wards, and emergency departments were visited. Deaths outside the hospital setting were tracked in the vital statistics records of the municipal health department.
In the analyses, multiple and singleton pregnancies were initially compared according to maternal characteristics and duration of the pregnancy within each cohort and between the two cohorts. In these analyses, the denominator was the number of mothers in each cohort.
Later, the rst-and second-born twins from the same cohort were compared to each other and to the singleton newborns according to birthweight, 5-minute Apgar score, NICU admissions, neonatal mortality rate (NNMR) (0-27th day of life), post-neonatal mortality rate (PNMR) (28th to 364th day of life), and IMR (0-364th day of life), all per 1,000 live births (LB). In these analyses, the denominator was the number of newborns in each cohort. Newborns from triplet pregnancies were excluded from these analyses. Means and standard deviations were calculated for the continuous variables, and proportions and 95% con dence intervals (95%CI) for the categorical variables. Student t-tests were used to compare the means, and chi-square tests to compare heterogeneity of categorical variables. The Weinberg method was used to estimate the number of monozygotic and dizygotic pairs.(28) According to this method, the dizygotic twin rate is double the rate of twins in which the pairs' sex is discordant, and the rate of monozygotic twin pregnancies is calculated as the difference between the total number of twin pregnancies and the dizygotic rate.

Results
Among the 4,187 pregnancies in the 2004 cohort, 42 (1.0%) were multiple; among the 4,220 pregnancies in the 2015 cohort, 56 (1.3%) were multiple, two of which were triplet pregnancies. Of the multiple pregnancies, 84 twins were born alive in 2004 and 111 in 2015. In 2004, the estimated rate of dizygotic pairs was 34%, and in 2015 it was 62%.
More than one third (67.9%) of the mothers in the 2004 cohort and more than half (59.9%) of those in the 2015 cohort belonged to families with monthly income ≤ 3 minimum wages ( Table 1). The proportion of women from families that earned one minimum wage or less decreased by half during the period (from 24.1% in 2004 to 12.7% in 2015; p < 0.001). There was a major change in education from 2004 to 2015, whereby the proportion of mothers with more schooling (12 years or more) tripled during the period (from 10.1-31.0%). As for age, in  None of the maternal sociodemographic characteristics or parity were associated with the occurrence of multiple pregnancies in either of the two cohorts (Table 2). In 2004, the highest twinning rates occurred at the extremes of family income (1.5% of births in families earning ≤ 1.0 minimum wage and in families earning > 10 times the monthly minimum wage) and among mothers with less than four years of schooling (1.8%). In 2015, the highest rates were in families with higher monthly incomes (2.8% and 1.9%, respectively, in families that earned 6.1-10 and > 10 minimum wage) and among mothers ≥ 35 years of age (2.3%). The majority of twin births were by cesarean section (78.6% in 2004 and 86.0% in 2015).  The mean twin-to-twin interval in the nine pairs of twins born by vaginal delivery in 2004 was 8.56 minutes (SD = 9.15). In the 2015 cohort, the mean twin-to-twin interval in the seven pairs of twins born by vaginal delivery was 24.14 minutes (SD = 32.67) (p-value of the difference between cohorts = 0.002). Table 4 compares second-born to rst-born twins, and these to the newborns of singleton pregnancies from the same cohort, according to the mean birthweight and unfavorable perinatal outcomes. There was no difference between rst-and second-born twins in the same cohort according to the target perinatal outcomes. In both cohorts, except for low 5-minute Apgar score, all the other indicators were more frequent in twins compared to singletons. Compared to the rst-and second-born twins in 2004, the NICU admissions rates in 2015 were higher, respectively, in rst-born (p = 0.005) and second-born twins (p = 0.04).   The study's limitations include the lack of information on family history of multiple pregnancies and the small number of twin births in both cohorts, which affects the study's power to identify associations between potential risk factors and the outcomes. The small sample size also prevented comparing perinatal outcomes and mortality between rst-and second-born twins according to gestational age and type of delivery, which are important prognostic characteristics for twins' survival, as reported by other authors. (11) Meanwhile, the use of two population-based birth cohorts from the same city, eleven years apart, with a low rate of attrition is a strength of this study. Data collected for each twin separately at the time of birth is also a strength mainly considering that information on second-born twins tends to be insu cient, especially when obtained from hospital medical records. (30) In addition, the mothers and twins were analyzed separately, and the rst-and second-born twins were analyzed individually according to incidence of unfavorable perinatal outcomes and mortality in the rst year of life. Although the prevalence and consequences of twin pregnancies may vary from place to place, the results of this study may be valid to other settings with similar socio-economic and maternal sociodemographic characteristics.
There was no difference in twinning rates between the two cohorts, but the rate in 2015 (1.4%) was higher than in the Brazilian state where Pelotas is located (Rio Grande do Sul) (1.2%) (p = 0.04) and in Brazil as a whole (1.13%) (p = 0.002) from 2011 to 2014.(4) This nding is consistent with the fact that the prevalence of some maternal factors associated with the occurrence of twin pregnancies, such as higher age and obesity, (20,31)  In both cohorts, there was no statistical difference in the multiple pregnancy rate according to maternal age or family income. However, in 2015, multiple pregnancies were more frequent in mothers 35 years or older (2.3%) and in families with monthly incomes between 6.1 and 10.0 minimum wage (2.8%). Older maternal age, together with genetic background and ethnicity, are known factors for multiple pregnancies. (1,20,30,34) As for income, the current nding is consistent with the results of previous analyses in this cohort. Eighteen mothers in 2015 gave birth to 23 newborns conceived by assisted reproduction procedures (in vitro fertilization in 70.6% of the cases). Nine of the 23 children were newborns from multiple pregnancies. (35) Although black ethnicity, multiparity, and low socioeconomic status have been associated with multiple pregnancies in other studies, particularly dizygotic twins,(4) we found no difference in these factors in our two cohorts. On the contrary, in the 2015 cohort, mothers with the highest family incomes showed higher than average twinning rates. In addition, a study that compared twinning rates in Pelotas from 1982 to 2015 showed a 220% increase among mothers with higher socioeconomic status and a 60% increase in white mothers, but no similar trend in brown or black mothers. (36) In our study, the small number of twin pregnancies may explain the lack of association between self-reported skin color and parity.
The estimated rate of dizygotic twins nearly doubled in Pelotas in the last decade. The factors associated with monozygotic and dizygotic twin pregnancies are different. While monozygotic pregnancies are mostly determined by genetic factors and in pregnancies resulting from in vitro fertilization, dizygotic pregnancies have more sociodemographic determinants, including ethnicity (more frequent in black women), multiparity, maternal age 35 years or older, low socioeconomic status, oral contraceptive use, family history, higher maternal height, obesity, and use of assisted reproduction techniques. (4,31) However, the method we used to estimate the type of twin pregnancy is more robust and reliable when applied to larger databases. (37,38) The increase in prematurity among twins differs from the relative stability in the overall prematurity rate in births in Pelotas during this period (13.7% in 2004 and 13.8% in 2015). (27) Multiple pregnancies are the principal factor independently associated with spontaneous preterm births in Brazil, (39) and twin pregnancies are also risk factors for medically induced preterm births. (40) In the two cohorts, all unfavorable perinatal outcomes were worse in second-born than rst-born twins. However, contrary to reports by other authors, (6,(11)(12)(13)(14) the differences were not statistically signi cant, probably due to the small sample size. Compared to singletons, the risk of second-born twins in the 2015 cohort dying in the post-neonatal period and in the rst year of life was, respectively, eight and six times greater, indicating their greater vulnerability.

Conclusion
This study showed that the prevalence of twin births remained stable in the eleven-year period, whereas the prevalence of births in multiple pregnancies prior to 34 weeks more than doubled from 2004 to 2015. Furthermore, in the 2015 cohort, the post-neonatal and the infant mortality rates in second-born twins were, respectively, eight and six times higher than in singletons. Given the representativeness of the sample (two population-based birth cohorts in the same city, using similar methodology, eleven years apart) these results could serve as a reference for the planning of health resources and development of preventive strategy targeting twin infants born in Pelotas and in settings with similar sociodemographic and health services structure characteristics. .5313). The mother or the child's legal guardian signed the free and informed consent form before the data were collected.

Consent for publication
Not applicable.

Availability of data and materials
The datasets analysed during the current study are available from the corresponding author on reasonable request.