Absolute Risk of Adverse Obstetric Outcomes Among Twin Pregnancies After In Vitro Fertilization by Maternal Age

Key Points Question What is the absolute risk of adverse obstetric outcomes stratified by in vitro fertilization (IVF), twin or singleton pregnancy, and maternal age? Findings In this cohort study of 16 879 728 pregnant women, the twin pregnancy rate was 32.1% among those who conceived via IVF. Twin pregnancies conceived via IVF had higher absolute obstetric risks in each maternal age compared with IVF-conceived singleton pregnancies or non–IVF-conceived twin pregnancies. Meaning These findings suggest that twin pregnancy, IVF, and advanced maternal age are independently associated with adverse obstetric outcomes, and their coexistence may lead to the aggravation of obstetric risk.


Introduction
In vitro fertilization (IVF) technologies have developed and spread globally during the past 4 decades since the first IVF-conceived infant was born in 1978. 1 Currently, IVF technologies mainly include IVF and embryo transfer, intracytoplasmic sperm injection, frozen embryo transfer, and preimplantation genetic testing, which have been widely applied among couples with infertility or monogenic diseases or with the intention of fertility preservation. 1,2 The updated global estimated number of IVFs per year was approximately 2.8 million initiated cycles and 0.9 million infants in 2012. 3 In China, there were 906 840 IVF cycles and 289 836 IVF infants in 2016, 4 accounting for 1.6% of the total 17.86 million births in the whole country in the same year. 5 Twin pregnancy is a common occurrence in pregnancies conceived with IVF because multiple embryo transfer is commonly regarded as an effective strategy to improve the likelihood of a successful pregnancy. 6 Globally, the twin delivery rates after IVF were 18.0% among fresh nondonor IVF embryo transfer and intracytoplasmic sperm injection cycles and 11.1% among frozen embryo transfer nondonor cycles in 2012, varying to some extent in different countries and regions. 3 In 2016, such twin delivery rates in Europe were 14.9% and 11.9%, respectively. 7 The rates in the US were 18.8% and 13.7%, respectively, 8 whereas twin delivery rates in China were 27.9% among fresh IVF embryo transfer cycles, 27.2% among fresh intracytoplasmic sperm injection cycles, and 24.2% among frozen embryo transfer cycles (egg and embryo donation are rarely performed in China). 4 Compared with singleton pregnancies, twin pregnancies are significantly linked with increased perinatal morbidity and mortality, including maternal near-death events, preterm birth, low birth weight, cesarean delivery, admission to the neonatal intensive care unit, stillbirth, and perinatal mortality. [9][10][11][12] These adverse obstetric outcomes are more likely to occur among pregnant women who underwent IVF conception than among those who conceived naturally. [13][14][15][16] Therefore, the risk of adverse obstetric outcomes among twin pregnancies after IVF may confoundingly originate from both twin pregnancy and IVF. In addition, maternal age is a critical independent factor for obstetric outcomes, whether after IVF or natural conception, [17][18][19] and presents nonlinear associations with many outcomes. 17 However, the risks of obstetric outcomes stratified by IVF, twin or singleton pregnancy, and maternal age are unknown. This study aimed to estimate the absolute risk of obstetric outcomes stratified by IVF or non-IVF conception and twin or singleton pregnancy at each maternal age to accurately evaluate the obstetric risks among twin pregnancies after IVF and then to develop management strategies in both IVF procedures and obstetric health care to ensure the health of mothers and infants.  (Figure 1).

Variable Definitions
The variables used in this study can be divided into 4 categories. Categories included maternal group, sociodemographic characteristics, maternal chronic diseases, and obstetric outcomes.

Maternal Group
In vitro fertilization was defined as pregnancy conceived with any technology of IVF. In vitro fertilization or non-IVF was identified according to the ICD-9-CM or ICD-10 codes of IVF procedures.
Singleton pregnancy or twin pregnancy was also identified by the related ICD-9-CM or ICD-10 codes.
The study population was then divided into 4 subgroups: singleton pregnancy with IVF (IVF-S), twin pregnancy with IVF (IVF-T), singleton pregnancy with non-IVF (nIVF-S), and twin pregnancy with non-IVF (nIVF-T).
Year was defined as the discharge date of the pregnant woman. Maternal age was defined as the pregnant woman's age at delivery (including either live birth or stillbirth). Ethnicity was defined as the ethnic group (Han or minority) to which the pregnant woman belonged.

Maternal Chronic Diseases
Maternal chronic diseases were defined as chronic diseases that the pregnant woman had before pregnancy, including chronic hypertension, diabetes, thyroid diseases, anemia, circulatory diseases, and other diseases (coagulation disorders, kidney diseases, diseases of connective tissues, diseases of the respiratory system, and diseases of the digestive system). All variables in this category were identified according to the ICD-9-CM or ICD-10 codes.

Obstetric Outcomes
Maternal outcomes were defined as maternal complications that developed during pregnancy, including gestational hypertension, eclampsia and preeclampsia, gestational diabetes, placenta previa, placental abruption, placenta accreta, preterm birth (gestational age at birth, 28-36 weeks), dystocia, cesarean delivery, and postpartum hemorrhage (bleeding volume Ն500 mL after vaginal delivery or Ն1000 mL after cesarean delivery). Neonatal outcomes were defined as neonatal complications that developed before or after birth until discharge, including fetal growth restriction (an estimated fetal weight during ultrasonographic screening that is less than the 10th percentile for gestational age), low birth weight (<2500 g), very low birth weight (<1500 g), macrosomia (birth weight >4000 g), malformation (congenital malformations, deformations, and chromosomal abnormalities), and stillbirth (the death or loss of an infant before or during delivery at 20 weeks of gestational age or later). Among these variables, low birth weight and very low birth weight were identified according to birth weights reported in the records, and other variables were identified according to the related ICD-9-CM or ICD-10 codes.

Statistical Analysis
Continuous variables are described as means (SD), and comparisons in different groups were performed using the 2-tailed t test. Categorical variables are described as counts with percentages, and comparisons in different groups were performed using the χ 2 test with relative risk (RR) and 95% CI. Poisson regression models using restricted cubic splines of maternal age were performed to examine the risk of each obstetric outcome by each maternal age in each subgroup. The level of significance was defined as P = .05, and all hypothesis tests were 2 sided. All analyses were conducted using SAS, version 9.0 software (SAS Institute, Inc). 24 Data were analyzed from September 1, 2020, to June 30, 2021.
In total, 16 obstetric outcomes were examined in this study, including 10 maternal complications and 6 neonatal complications. For each outcome, we performed 3 Poisson regression models: (1) using only singleton pregnancy data to examine the adjusted RR (aRR) and 95% CI of IVF-S vs nIVF-S (model 1), (2) using only twin pregnancy data to examine the aRR and 95% CI of IVF-T vs nIVF-T (model 2), and (3) using both singleton pregnancy and twin pregnancy data to examine the aRR and 95% CI of IVF vs non-IVF (model 3). Geographic region, maternal age, year, ethnicity, and maternal chronic diseases were adjusted for in all models, and singleton or twin pregnancy was additionally adjusted for in model 3. The interaction effect between IVF and twin pregnancy was tested by including an interaction term in model 3. Maternal age was modeled using restricted cubic splines to allow a nonlinear association with each outcome, and the number of knots placed at the default percentiles was determined according to the principle of minimized Akaike information criterion in the adjusted model. 25,26 In model 3, we computed estimated absolute risks (probabilities and 95% CIs) in each subgroup and then presented them graphically to visually illustrate the risk of each obstetric outcome by each maternal age.

Sensitivity Analyses
The    between IVF and twin pregnancy for maternal outcomes except for dystocia (eTable 6 in the Supplement). Further analyses of twin pregnancies showed that the IVF-T group had a higher risk of 9 outcomes than the nIVF-T group, but there was no difference in placental abruption (aOR, 1.03; 95% CI, 0.96-1.10). Notably, the rate of cesarean delivery was fairly high in each subgroup: 88.8% in the IVF-T group, 80.3% in the nIVF-T group, 66.5% in the IVF-S group, and 43.6% in the nIVF-S group.
f The value of the aRR is significantly lower than 1.00.
g Among a total of 16 879 728 pregnant women aged 20 to 49 years analyzed in this study, there were 1 879 110 women (11.1%) with missing values in birth weight who were not included when performing the description and comparison of low birth weight and very low birth weight.  Figure 2, accompanied by the eFigure and eTable 9 in the Supplement, presents the estimated absolute risks (probabilities and 95% CIs) of each outcome in each subgroup at each maternal age ranging from 20 to 49 years. These curves show that the obstetric risk in each subgroup was almost always elevated with increasing maternal age. Although the forms of each curve differed, they can be summarized in 2 dominant patterns. Pattern A indicated the absolute risk ranging from IVF-T to nIVF-T to IVF-S to nIVF-S, which presented with gestational hypertension, eclampsia and preeclampsia, placental abruption, preterm birth, dystocia, cesarean delivery, postpartum hemorrhage, fetal growth restriction, low birth weight, very low birth weight, and malformation.

Absolute Obstetric Risk by Maternal Age, IVF, and Twin Pregnancy
Pattern B indicated the absolute risk as ranging from IVF-T to IVF-S to nIVF-T to nIVF-S, which presented with gestational diabetes and placenta accreta. The absolute risk of placenta previa partly complied with pattern B, but there was no difference between singleton and twin pregnancies in the IVF group or non-IVF group at certain intervals of maternal age. In addition, there were 2 sporadic patterns: singleton pregnancy had a higher risk of macrosomia than twin pregnancy, but there was no difference between the IVF group and the non-IVF group; and the risk of stillbirth ranged from nIVF-T to IVF-T to nIVF-S to IVF-S.

Discussion
Based on national hospital data, we estimated the absolute risk of obstetric outcomes stratified by IVF or non-IVF conception and twin or singleton pregnancy at each maternal age using nonlinear models and found visual evidence suggesting that IVF-conceived twin pregnancies had a higher risk of maternal and neonatal complications than IVF-conceived singleton pregnancies or non-IVFconceived twin pregnancies. Our findings provide valuable evidence for clinical decision-making and public health policies.
Among IVF-conceived twin pregnancies, the most common obstetric risks were cesarean delivery, low birth weight, preterm birth, gestational diabetes, gestational hypertension, preeclampsia and eclampsia, dystocia, and postpartum hemorrhage, which is consistent with findings of previous studies conducted in varied countries and populations. 6,13,19 Most curves of obstetric risks showed an elevated trend with increasing maternal age, especially among women older than 35 years, which is also in line with findings of previous literature. 17,18,27,28 In this study, we further summarized the forms of each obstetric risk curve into 2 dominant patterns. Pattern A presented that at each maternal age, the risk effects of twin pregnancy appeared more prominent than those of IVF, whereas pattern B presented the opposite appearance. The former was more common than the latter, but both revealed that IVF-conceived twin pregnancies had the highest risks in most obstetric outcomes observed in this study. These findings indicate that twin pregnancy, IVF, and advanced maternal age (often defined as Ն35 years 29,30 ) are independent indicators for many adverse obstetric outcomes, and their coexistence will lead to the aggravation of obstetric risk.
In addition, 2 sporadic patterns of obstetric risks were present in macrosomia and stillbirth. The absolute risk of macrosomia appeared higher in singleton pregnancies than in twin pregnancies but was not different between the non-IVF group and the IVF group, which is in accordance with previous research. 31,32 The absolute risk of stillbirth appeared higher in twin pregnancies than in singleton pregnancies, which is also in line with previous literature 9,11 ; however, it appeared lower in the IVF group than in the non-IVF group, which differs from previous literature. 31,33 This may be due to a high compliance with antenatal examinations among pregnant women with IVF conception, especially for noninvasive prenatal testing or amniocentesis. 34,35 Because artificial abortion is legal in China, pregnancies with abnormalities may be terminated early, which can reduce the incidence of stillbirth reported after gestational age of 20 weeks.
One serious problem revealed in this study was the high rate of twin pregnancy among pregnant women in the IVF group (32.1%). With improvements in IVF technology and increasing recognition of the risk of multiple pregnancies, elective single-embryo transfer (eSET) is recognized as an effective strategy to reduce multiple pregnancies after IVF. 36,37 For instance, the UK introduced a policy in 2009 to encourage fertility centers to adopt eSET into routine use, which resulted in a sustained trend in the reduction of the multiple pregnancy rate from 26.6% in 2008 to 16.3% in 2013 38 ; it then dropped to 6% in 2019. 39 The US published eSET guidance for patient selection in 2012, recommending that eSET should be offered to patients younger than 35 years, 38 which led to a persistent reduction in the twin pregnancy rate in the group younger than 35 years from 30.8% in 2011 to 9.9% in 2018. 40 However, there is no legal regulation or clinical guideline to promote the use of eSET in China; thus, many assisted reproductive centers are not actively promoting eSET, which results in a higher rate of twin pregnancy along with the adverse pregnancy outcomes as mentioned above.
Another noteworthy problem revealed in this study was the high rate of cesarean delivery in each subgroup (IVF-T, 88.8%; nIVF-T, 80.3%; IVF-S, 66.5%; nIVF-S, 43.6%), which was much higher than the global average (21.1%) 41 or the upper threshold (15%-19%) recommended to reduce maternal and neonatal mortality. 42,43 Neither IVF nor twin pregnancy themselves are associated with cesarean delivery, if there are no serious complications medically indicating the necessity of cesarean delivery. [44][45][46] In China, the overuse of cesarean delivery has been driven by complex factors, such as the economic incentives for clinicians, a culture of obstetrician-led delivery, inadequate resources for pain relief in labor, and maternal request with inadequate risk awareness. 47,48 Therefore, unnecessary cesarean delivery is a major challenge in China, which can cause avoidable harms or unnecessarily increase the need for additional intervention and costs. 47 There is an urgent need to reduce unnecessary cesarean deliveries to improve maternal and offspring health.

Limitations
Several limitations must be considered when interpreting the findings of this study. First, there may be selection bias in our study, because pregnant women with serious obstetric complications or higher socioeconomic status are more likely to give birth in tertiary hospitals; thus, it is necessary to further expand the sampling hospitals of the HQMS to different levels with different population groups. Moreover, several crucial factors of obstetric risk were absent in the database of HQMS, such as parity, gestational age, the adoption of risk behaviors (eg, smoking or drinking), paternal characteristics, family history, the reason for cesarean delivery (with any medical indicator or not), the number of embryos transferred and oocyte source in IVF pregnancies, and the chorionicity and amnioticity of the twins; thus, we recommend establishing integrated health information systems for cross-sectoral sharing of data and link them with other medical databases or population databases to retrieve more available and reliable data for promoting evidence-based practices.

Conclusions
This cohort study estimated the absolute risk of obstetric outcomes at each maternal age among twin pregnancies conceived with IVF in a large Chinese population. The findings suggest that twin pregnancy, IVF, and advanced maternal age are independent indicators of adverse obstetric outcomes, and their coexistence may lead to the aggravation of obstetric risk. We call for developing comprehensive and evidence-based guidelines of health care management for pregnant women with such obstetric characteristics. Furthermore, there is a need to promote eSET to reduce multiple pregnancies after IVF, and unnecessary cesarean delivery should be avoided in all pregnant women.