Perinatal outcomes in twin late preterm pregnancies: results from an Italian area-based, prospective cohort study

Late preterm (LP) birth, defined as delivery between 34 and 0/7 and 36–6/7 weeks of gestation, accounts for about 70% of overall prematurity in Europe [1]. Compared to the term neonates, LP more commonly encounter adverse outcomes, such as higher respiratory morbidity [2‐4], long-term adverse developmental and cognitive outcomes [5] and may also have repercussions on their health as adults [6, 7].

Multiple gestations represent a considerable proportion of pregnancies delivering in the LP period. Twin gestations account for approximately 3–4% of all live births [8, 9], with a LP rate of 49.8% [10], and a mean gestational age (GA) at delivery of 35.3 weeks [11]. When compared to singleton gestations, twin mothers have higher morbidity and mortality rates from gestational hypertension, cesarean delivery, and postpartum hemorrhage, while their newborns have higher NICU admission rates and neonatal sequelae [12]. Furthermore, after adjusting for gestational age at delivery, monochorionic twins (MC) experience higher rates of perinatal complications and stillbirth (SB) than dichorionic twins (DT) and singletons [13‐15].

While 70% of all twin births are indicated, 30% are due to spontaneous preterm labor [16]. Despite guidelines indicating when to deliver twin gestations according to gestational age, chorionicity, and concomitant maternal or fetal complications [17], 16–50% of indicated LP twin deliveries are non-evidence based (EB) [17]. As non EB delivery indications account for iatrogenic morbidity that could be prevented, this observational study intended to investigate first neonatal outcomes of LP twin pregnancies according to GA at delivery, chorionicity and delivery indication, then non EB delivery indications.

We conducted a prospective, multicenter, area-based cohort study of the LP deliveries from December 1st 2013, to December 1st 2015 in Emilia Romagna, a region in Northern Italy with nearly 4.5 million people organized in 9 counties. The results of the single pregnancy cohort study were just published [18].

The study steering committee (consisting in 2 obstetricians, 2 neonatologists and 1 epidemiologist) defined the information to collect on pregnancies resulting in LP deliveries, and on neonates’ hospital course. Standardized chart abstraction forms were designed to obtain anonymized data on mothers (i.e maternal demographics, maternal medical and obstetrical complications, labor, and delivery details) and their newborns (i.e birthweight, gender, Apgar scores, admission to the NICU or Intermediate/Step Down Unit, and length of stay, neonatal morbidities, and mortality). In Emilia Romagna, inpatient obstetric and pediatric care was provided at the time of the study by 28 hospitals within the National Health System. Recruitment occurred only at the 21 hospitals with at least 500 deliveries/year, equipped with NICUs or Intermediate/Step Down Units, that could care for LP infants: one obstetrician and one neonatologist/pediatrician from each study site approached mothers with twin pregnancy delivering at 34^+ 0–36^+ 6 weeks for written informed consent, and respectively completed the maternal and the neonatal data collection forms, as they interviewed the patients and consulted the medical records. Five trained research associates visited the study sites monthly basis to review the maternal and neonatal data collection forms with the physicians that had filled them. At the Data Managing Center (Modena Policlinico Hospital, University of Modena, and Reggio Emilia) research associates used optical character recognition technology (Flexi Capture, Abbyy®, Germany) to scan the data collection sheets, extract pertinent data, and organize it in a password protected database. If automated data extraction failed (approximately 5% of the times), research associates reviewed the forms and manually entered the missing data in the database. Authors have complied with the World Medical Association Declaration of Helsinki regarding ethical conduct of research involving human subjects. Approval from the Institution Review Board of the 9 Emilia Romagna Counties was obtained (n 24,512 _ 25/9/2014). The study was financed by the Emilia Romagna County Grant (n 417,149 _ 2014).

Among the 346 LP twin gestations included in our cohort, 84 (23.4%) were monochorionic pregnancies (of which only 3 were MCMA), while 262 (75.7%) were dichorionic; sPTL accounted for 85 (24.6%) deliveries, pPROM for 66 (19.1%), EB indicated deliveries were 117 (33.8%), while non EB indications were 78 (22.5%).

In the univariate analysis shown in Table 1, maternal and obstetrics characteristics are presented according to delivery indication. When compared to sPTL or pPROM, pregnancies delivered due to maternal and/or fetal indications were associated with higher maternal age (p <  0.01), higher gestational age at delivery (p <  0.01), Caucasian race (p 0.04), ART use (p <  0.01), gestational diabetes (p <  0.01), vaginal bleeding (p <  0.01), prescription of antenatal corticosteroids (p <  0.01), diagnosis of FGR (p <  0.01), and were more frequently monochorionic (p <  0.01). Only 15.3% (53/546) of pregnancies had a vaginal delivery of both fetuses. Cesarean deliveries in non-laboring women were 204 (58.9%), induced labor 37 (10.7%), while sPTL were 105 (30.4%). Women experiencing sPTL had the highest rate of vaginal delivery (27.1%).

Table 2 summarizes neonatal characteristics. Two hundred twenty-six pregnancies (65.3%) had at least one fetus experiencing one adverse outcome; pregnancies with sPTL, pPROM and indicated deliveries had similar rates of the composite adverse neonatal outcome. Newborn characteristics were similar among study groups; however, mothers with indicated deliveries had lower birthweights (p <  0.01) and higher SGA rates (p <  0.01). Median NICU length of stay was similar among study groups, although stay ≥5 days was less common when indicated deliveries were non EB (p <  0.05).

Table 3 lists the EB indicated delivery (117, 33.8%) and non EB indicated deliveries (78, 22.5%) according to our expert review.

As detailed in Table 4, multivariate analysis confirmed that delivery indication did not affect the composite adverse neonatal outcomes; the only characteristic that continued to affect the outcome after controlling for confounding was gestational age at delivery (p <  0.01). Moreover, there was at least one adverse neonatal outcome for 94% of babies born at 34 weeks, for 73% of those born at 35 weeks and for 46% of those born at 36 weeks, with a significant difference (p <  0.01).

Although outcomes of singleton gestations at 34^+ 0–36^+ 6 were shown to vary according to delivery indications [37], our large prospective cohort of twin gestation did not corroborate such finding. Instead, we confirmed on twins that significant differences in neonatal outcomes depend on gestational age at delivery, as it was previously established for singletons [38]. Furthermore, we showed that almost 1 out of 4 twin gestation in our cohort was delivered due non EB indications.

The high prevalence of non EB delivery indications in our twin population raises great concerns. Twin pregnancies represent a fragile obstetric population, having worse outcomes than singletons [38]; furthermore, LP deliveries are associated with worse outcomes than early term and term deliveries [10, 39]. Therefore, lack of sound indications for twin deliveries in the LP window combines the downsides of prematurity with the ones of multiple gestations, multiplying the risks of severe complications. Our findings urge a critical review of clinical practice encouraging obstetric providers to familiarize with the indications that prompt delivery among twins, overcome the uneasiness associated with management of twin pregnancies, in order to prevent iatrogenic morbidity. In our cohort 10 deliveries (12.8%) were due to gestational hypertension, 15 (19.2%) to class A1 gestational diabetes mellitus (diet therapy), 2 (2.6%) to mild cholestasis of pregnancy, while no clear indication was stated in 29 instances (37.2%) and translate into 78 (22.5%) early deliveries that could be avoided.

Our results are consistent with previous reports on singleton gestations [40] showing higher rates of non-indicated deliveries in the LP window. Moreover, Reddy et al. reviewed delivery indications for almost 3.5 million births in the United States concluding that 23% of LP births had no documented indication: the author speculated that many of these births without listed indications were due to perceptions of impending morbidity and mortality risks by either patients or providers [41].

The balance between the downsides of prematurity and the risks associated with pregnancy continuation, can be hard to strike in the LP period, especially among twin gestations [31] accounting for a discrepancy in clinical practice according to hospital settings and countries [42]. A recent Delphy survey on delivery indications in singleton pregnancies, showed that Italian experts generally agreed to be more conservative at 34 weeks, favoring delivery at 36 weeks, whatever the clinical scenery [32]. Nevertheless, in our twin population the most non EB indicated deliveries (22.5%) were without a clear medical cause, maybe due to different clinical setting and provider advisability.

Although women with EB deliveries were more likely to have babies with adverse perinatal outcome, this did not remain significant in the multivariate analysis, where the only factor affecting neonatal morbidity was the gestational age at delivery, as demonstrated by other studies [43]. Delivery indication did not significantly affect the outcome of interest even after excluding from the analysis non EB indicated delivery (data not shown).

Six stillbirths were detected in our population: 5 among BC pregnancies, 1 among MC twins. Our stillbirth rate (1.7%) is consistent with the current literature; however, we did not find any difference according to chorionicity [15]. Furthermore, chorionicity did not affect the composite adverse neonatal outcome as shown in the multivariate analysis. Previous reports have shown greater neonatal morbidity and mortality among MC pregnancies, from placental vascular complications [44] such as in TTTS, or in case of intravascular shunting after a single intrauterine death; or due to congenital malformations, selective growth restriction, or maternal complications [13, 14, 45]. We can speculate that improved outcomes among MC twins could be attributed not only to an evidence based antenatal care [12, 46] but also to a generally healthier obstetric population, where risk factors such as obesity [47], hypertensive disorders and diabetes have relatively low prevalence.

The joined 2020 NIH [48] and SMFM consensus recommended delivery at 38 weeks’ gestation for uncomplicated DC twin pregnancies as opposed to 34–37 weeks’ gestation for uncomplicated MC twins [12]. This broad interval was recently revised by a retrospective cohort study showing that elective LP delivery of MC is associated with significantly increased neonatal morbidity that does not seem justifiable by a corresponding reduction in the risk of SB. Therefore, the authors concluded that with a specialized antenatal care for twins and intensive fetal surveillance, delivery at 37 weeks’ gestation appears to be associated with the best outcomes for uncomplicated MC [13]. Our data confirm such findings, suggesting to avoid LP birth both in uncomplicated MC and DC twins to reduce unnecessary admission to NICU and other neonatal complications.

Strengths of this study include the large sample size, area based and multicenter nature of the cohort: both characteristics increase the generalizability of our findings. The prospective design of the survey, along with predefinition of standardized chart abstraction forms completed by obstetricians and pediatricians, and periodically audited by research associates, limit misclassification bias and assures data validity. We also acknowledge some limitations. As the study did not provide clinicians with prespecified delivery indications, it could be challenging at times for study personnel to identify the rationale behind an indicated delivery. However, such lack of standardization also allowed us to collect information about indications that were non EB. Although inconsistencies in ANCS administration may reflect different opinion leaders’ viewpoints, we considered betamethasone treatment among the potential confounders in our multivariate analyses. Moreover, our data are related to the years 2013–2015 and clinical practices might have slightly changed in these years. Finally, due to the exploratory nature of our study, a specific sample size calculation was not performed.

To our knowledge, this is one of the largest prospective studies investigating delivery indications among twin gestations, suggesting that the decision to deliver or not twins in the LP period should consider gestational age at delivery as the main determinant infants’ prognosis. Delivery indications should therefore be accurately considered, to avoid iatrogenic early deliveries responsible of preventable complications. Our findings can also be helpful when counselling mothers at risk of LP delivery and can be used to plan interventions for their newborns.