Impact of antenatal corticosteroid therapy on neonatal outcomes in twin pregnancies

Abstract

Background

Corticosteroids are widely used in obstetric clinical practice for cases with signs of preterm labor to promote fetal lung maturity and reduce neonatal morbidity and mortality. Although short-term use is considered safe, there is ongoing debate regarding the dosage, therapeutic window, neonatal benefits, and maternal-fetal side effects, especially in high-risk pregnancies such as twins, where the impact remains unclear.

Methods

This retrospective study included 1,997 twin pregnancies, divided into two groups: those who received antenatal corticosteroid therapy (ACS) and those who did not. To correct for baseline imbalances, the optimal overlap weighting scheme was selected by calculating the Absolute Standardized Mean Difference (ASMD) to minimize intergroup differences. The primary outcome, neonatal respiratory distress syndrome (NRDS), and other adverse outcomes in twin neonates were analyzed for the effect of ACS using logistic regression, with subgroup and interaction analyses based on key maternal pregnancy characteristics. Lastly, the Restricted cubic spline (RCS) method was used to examine the effect of ACS on neonatal respiratory disease incidence across different gestational ages at delivery.

Results

After propensity score overlap weighting, results showed that although ACS treatment did not significantly improve the respiratory composite outcome in the overall preterm group, it effectively reduced the incidence of NRDS and pneumonia, while also decreasing the risk of low birth weight, small for gestational age (SGA), neonatal purpura, and neonatal hypoproteinemia. Notably, the risk of neonatal hypoglycemia and hyperbilirubinemia was significantly increased in the ACS treatment group. In both early and late preterm groups, there was no significant difference in the impact of ACS on NRDS and respiratory composite outcomes, but it remained effective in reducing the risks of neonatal pneumonia, low birth weight, and hypoproteinemia. In late preterm pregnancies, ACS significantly reduced the incidence of neonatal enteritis, lower gastrointestinal bleeding and neonatal infections, while in early preterm pregnancies, it significantly lowered the risk of neonatal hyperlacticemia. Subgroup analysis showed that for early preterm twin pregnancies with gestational diabetes mellitus (GDM), ACS treatment increased the incidence of NRDS and the neonatal respiratory composite outcome. Similarly, for twin pregnancies complicated by preeclampsia (PE), ACS treatment raised NRDS incidence in both overall and early preterm subgroups. Finally, RCS analysis indicated that ACS treatment may help reduce the risk of NRDS and other respiratory outcomes across different gestational ages at delivery, although this trend did not reach statistical significance. Sensitivity analysis showed similar results.

Conclusion

Antenatal corticosteroids, whether in early or late preterm births, may not prevent NRDS and respiratory composite outcomes in twin neonates, but they are effective in reducing adverse neonatal outcomes such as pneumonia, low birth weight, and hypoproteinemia. However, the occurrence of neonatal hypoglycemia and hyperbilirubinemia should be noted.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-06679-w.

Highlights

No impact of ACS on NRDS ACS did not significantly reduce NRDS or respiratory composite outcomes in twin pregnancies, regardless of gestational age.

Reduction in neonatal morbidity with ACS ACS significantly reduced neonatal pneumonia, low birth weight, SGA, hypoproteinemia, and in late preterm pregnancies, also reduced enteritis, gastrointestinal bleeding, and infections.

Maternal conditions modulate ACS effectiveness ACS increased the risk of NRDS in early preterm twins with GDM or PE.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-06679-w.

Introduction

Over the past four decades, the incidence of twin pregnancies has significantly increased on a global scale. Concurrently, the rate of preterm birth in twin pregnancies has remained high [1]. Data from the World Health Organization’s multinational study on maternal and neonatal health indicate that the rates of preterm (28 to 36 weeks, 37.1%) and very preterm births (28 to 33 weeks, 13.0%) in twin pregnancies are significantly higher than those in singleton pregnancies (preterm rate 7.3%, very preterm rate 2.4%), leading to increased morbidity and mortality and posing a significant societal burden [2].

For singleton pregnancies at risk of preterm birth, the use of ACS has been proven to effectively reduce neonatal morbidity and mortality. However, the efficacy of ACS in twin pregnancies, a high-risk group for preterm birth, remains uncertain. Currently, there is a limited number of randomized trials on the effects of ACS in multiple pregnancies. Although pooled risk ratios suggest that ACS treatment may be beneficial in multiple pregnancies, there is considerable uncertainty surrounding these results, necessitating further research to confirm the actual effects of this treatment [3, 4]. Additionally, twin pregnancies are associated with higher morbidity compared to singletons, which may affect the use and evaluation of ACS. For example, a previous Korean nationwide cohort study showed that twins had a higher risk of high-grade intraventricular hemorrhage (IVH) and retinopathy of prematurity (ROP) in the 23–28 weeks gestation group, and a higher risk of periventricular leukomalacia (PVL) in the 29–33 weeks group. These differences suggest that the impact of ACS could vary between twins and singletons, complicating its assessment and effectiveness in this population [5]. A recent meta-analysis explored the effects of ACS in twins [6]. Despite reports that ACS reduce the risk of adverse neonatal outcomes, the credibility of these findings is limited due to potential confounding factors in observational studies of drug efficacy [7, 8].

Regarding the timing window for ACS administration, high-quality singleton RCTs support the use of a single course of ACS in women at risk of preterm birth between 24 0/7 and 33 6/7 weeks of gestation to prevent NRDS and reduce neonatal mortality [9]. Some studies specifically on twin pregnancies show similar effects of ACS in preventing RDS and reducing neonatal mortality compared to singletons [10]. However, larger-scale cohort studies in twin pregnancies are needed to further confirm the efficacy of ACS within this window. For late preterm pregnancies (34 0/7 to 36 6/7 weeks), although ACS has shown positive effects in reducing respiratory complications in singleton neonates, its impact on twins remains unclear. Early studies suggest that ACS treatment for late preterm birth in twins does not significantly reduce neonatal respiratory morbidity and may be associated with a higher risk of neonatal intensive care unit (NICU) admission and hypoglycemia [11–13]. Therefore, the impact of ACS treatment on neonatal outcomes in twins at risk of late preterm birth remains an unresolved issue.

To further explore the impact of ACS treatment on neonatal outcomes in twin pregnancies at risk of preterm birth, this retrospective cohort study included 1,997 twin pregnancies. It systematically examined the effects of ACS on neonatal outcomes in early preterm, late preterm, and overall preterm twin pregnancies, offering valuable insights for the rational use of ACS in obstetric practice for twin pregnancies.

Methods

Ethical approval

This study was approved by the Women and Children’s Hospital of Chongqing Medical University (ID: 2022-011-01). Informed consent to participate was obtained verbally from all study participants. This method of consent was deemed appropriate due to the nature of the study, where minimal risk to participants was involved, and written consent was not deemed necessary. The use of verbal consent was approved by the ethics committee of the Women and Children’s Hospital of Chongqing Medical University. To safeguard patient privacy, all personally identifiable information was removed from the cases, and all acquired data were kept anonymous.

Study design

This study divided twin pregnancies into two groups based on whether they received ACS, and analyzed them separately in all preterm, early preterm, and late preterm pregnancies. After adjusting for baseline differences using various matching methods, the study assessed the effect of ACS on neonatal adverse outcomes, including the primary outcome of NRDS.

Setting

This retrospective cohort study included all twin pregnancies that were managed at the Women’s and Children’s Hospital of Chongqing Medical University between Month Day, 2017, and Month Day, 2022. The hospital is a tertiary care center. These individuals were identified through the institution’s electronic medical record system, from which electronic records were extracted to collect data on maternal characteristics, pregnancy outcomes, neonatal outcomes, and ACS treatment. Gestational age (GA) was calculated based on the last menstrual period, adjusted by the crown-rump length in the first trimester or the head circumference of the larger fetus after 14 weeks of gestation as determined by ultrasound, and the dates of in vitro fertilization pregnancies conceived through assisted reproductive technology. Our study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for cohort studies.

Exposure

Exposure was defined as having received at least one dose of antenatal corticosteroids prior to 36 weeks and 6 days. The antenatal corticosteroid treatment protocol included four doses of intramuscular dexamethasone, each dose 5 mg, administered every 12 h. High probability of delivery was defined as spontaneous rupture of membranes, intact amniotic sac with at least 3 cm dilation or 75% cervical effacement, or the expectation of inducing labor or performing cesarean delivery within 24 h to 7 days for any other reason [14]. The data included information on whether corticosteroids were administered during the mother’s hospital admission for delivery, but did not include the type of medication, number of doses, or whether corticosteroids had been given during a previous admission. Post-delivery, maternal characteristics and neonatal outcomes of twins were compared between the two study groups. The exposure settings in this study are consistent with previous research [12].

Data collection

All data for this study were sourced from the electronic medical records (EMR) database of the Women and Children’s Hospital of Chongqing Medical University. The data were collected by trained medical staff, who reviewed the medical records to extract basic information about pregnant women and pregnancy-related data. The database was constructed by systematically compiling patient information from electronic records, ensuring comprehensive inclusion of detailed demographic and clinical characteristics.

Our outcomes and variables were primarily based on strict pathological or clinical diagnoses, which were further validated through ICD code verification. We did not rely solely on ICD codes for diagnosis, but used them in conjunction with clinical assessments to ensure the accuracy of the diagnoses. This dual approach significantly enhances the reliability of our study population’s diagnoses and mitigates the potential risks associated with mechanical extraction of ICD codes. The diagnostic criteria used for data inclusion were clearly defined and consistently applied throughout the review process.

Participants

The study retrospectively included pregnant women who received regular prenatal care and delivered between 2017 and 2022. Inclusion criteria were as follows: (i) complete medical records (12% of the medical records were incomplete); (ii) twin pregnancies; (iii) gestational age ≥ 24 weeks. Exclusion criteria were: (i) GA at delivery greater than 37 weeks; (ii) severe congenital or chromosomal fetal abnormalities (45 cases); (iii) cases of fetal demise in utero (63 cases), with exclusion of both newborns in multiple pregnancies if one fetus is affected.; (iv) complications unique to monochorionic twins, including twin-to-twin transfusion syndrome (TTTS), twin reversed arterial perfusion sequence, or twin anemia-polycythemia sequence (38 cases); (v) cases missing outcomes due to transfer at birth without respiratory complications (17 cases); (vi) chorioamnionitis or maternal fever at delivery (57 cases). The inclusion and exclusion criteria are consistent with previous studies [12, 15]. A total of 2,154 women with twin pregnancies met the inclusion criteria. After applying the exclusion criteria, 1,997 women with twin pregnancies and 3,994 neonates were ultimately included. The study flowchart is shown in Fig. 1.

Fig. 1.

Flowchart of this retrospective cohort study

Variables

Demographics and pregnancy characteristics

We summarized characteristics potentially associated with neonatal respiratory outcomes. Pre-pregnancy characteristics included maternal age (years), body mass index (BMI in kg/m²), parity (nulliparous vs. multiparous), gravidity (primigravida vs. multigravida), and family history (hypertension, diabetes, cancer). Pregnancy characteristics encompassed chorionicity (DCDA vs. non-DCDA), use of assisted reproductive technology (ART), scarred uterus, uterine myoma, placenta previa, placenta accreta spectrum, pre-eclampsia (PE), gestational diabetes mellitus (GDM), intrahepatic cholestasis of pregnancy (ICP), and gestational hypothyroidism, gestational anaemia, gestational thrombocytopenia. Definitions of characteristics are provided in Supplementary Table 1.

Neonatal outcomes

Preterm birth (PTB) is defined as childbirth occurring before 37 completed weeks of gestation. For the purposes of this study, PTB is further classified into two categories: early PTB, defined as birth occurring between 24 weeks 0 days and 33 weeks 6 days of gestation, and late PTB, defined as birth occurring between 34 weeks 0 days and 36 weeks 6 days of gestation [16]. The primary outcome measure was neonatal respiratory distress syndrome (NRDS), diagnosed based on progressive clinical symptoms (tachypnea, nasal flaring, retractions, grunting, and cyanosis), arterial blood gas analysis (hypoxemia, elevated CO₂, and acidosis), and typical chest X-ray findings (fine granular densities, air bronchogram, ground-glass opacity, or “white lung” appearance in severe cases). Secondary outcomes included the following respiratory outcomes: neonatal respiratory failure (NRF), neonatal pneumonia, and neonatal respiratory composite outcome, defined as at least one of the following occurrences: NRDS, NRF, or neonatal pneumonia. Digestive outcomes included neonatal enteritis, lower gastrointestinal hemorrhage, and neonatal digestive composite outcome, defined as at least one of the following occurrences: neonatal enteritis or lower gastrointestinal hemorrhage. Other outcomes included neonatal infection, LBW (low birth weight), SGA (small for gestational age), polyhydramnios, oligohydramnios, neonatal purpura, neonatal hypoglycemia, neonatal hyperbilirubinemia, neonatal hypoproteinemia, and neonatal hyperlactatemia. Definitions of neonatal outcomes are provided in Supplementary Table 2.

Statistical methods

Statistical analyses were conducted using SPSS software version 26.0 and R Studio version 4.0.2. Participants were initially categorized based on whether they received ACS. To describe the sociodemographic characteristics of participants, normally distributed continuous data were presented as means and standard deviations, while non-normally distributed data were reported using medians and interquartile ranges (IQR).

Given the risk of bias in traditional non-randomized studies due to confounding factors, as pregnancies treated with antenatal corticosteroids may have different clinical risk profiles compared to untreated pregnancies, it is essential to use statistical methods to adjust for these differences and ensure more reliable results. Adjusting for measured confounders, such as through multivariable regression, can help mitigate some of this bias, but it is rarely possible to eliminate all differences in risk profiles between groups [12, 17]. To adjust for intergroup differences, this study developed propensity scores to reflect the probability of each participant receiving antenatal corticosteroid treatment, using logistic regression with exposure (antenatal corticosteroid treatment vs. no antenatal corticosteroid treatment) as the dependent variable and all observed maternal-level characteristics as independent variables [18]. Covariates encompassed baseline characteristics of the pregnant women, including sociodemographic and clinical factors, which were anticipated to be related to neonatal outcomes.

Given the different baseline characteristics between total, early, and late preterm births, propensity scores were developed separately for each group. For early preterm births, variables such as placental abruption, placenta previa, and hypertensive disorders of pregnancy (HDP) are stronger predictors, while gestational diabetes mellitus (GDM) and intrahepatic cholestasis of pregnancy (ICP) are more relevant for late preterm births. Additionally, maternal factors like age under 24 years are associated with early preterm birth, but not with late preterm birth. In contrast, factors like obesity is linked to increased late preterm birth risk. These differences underscore the need for separate propensity score models [19, 20]. Based on the propensity scores, three adjustment methods were employed to match the baseline characteristics of the two groups receiving and not receiving antenatal corticosteroid treatment: Propensity Score Matching (PSM), Inverse Probability of Treatment Weighting (IPTW), and Overlap Weighting (OW). The association between ACS and neonatal outcomes was examined using OW and a 1:1 PSM approach with nearest-neighbor matching (caliper width of 0.1), to check for robustness. The best adjustment method was selected based on the Average Standardized Mean Difference (ASMD), a statistical measure used to assess the balance of covariates between groups in observational studies. Among the different matching methods, 1:1 PSM (Propensity Score Matching) showed poorer stability. In contrast, both OW and ITPW resulted in ASMD values of less than 10% for all covariates. Notably, the OW method demonstrated the best balance across all three matching iterations, supporting the hypothesis of good balance between the study groups. OW was chosen for subsequent analysis as all covariates had an ASMD of less than 10%, supporting the hypothesis of balance between the study groups [21] (Supplementary Figs. 1–3). It is important to note that although the matching model in this study was based on the baseline characteristics of twin pregnancy mothers, each child was still analyzed individually. In addition, since all analyses conducted in this study are exploratory, no adjustments were made for multiple testing [22].

Subgroup analyses and interaction effects

This study performed six subgroup analyses to examine heterogeneity in clinically relevant groups for NRDS and the neonatal respiratory composite outcome: chorionicity (DCDA versus non-DCDA), use of assisted reproductive technology (ART), parity (nulliparous versus multiparous), gestational diabetes mellitus (GDM), intrahepatic cholestasis of pregnancy (ICP), and preeclampsia (PE). The effect of ACS on the primary outcome within each subgroup was first assessed using a generalized linear model (GLM) with a logistic regression link function, as the primary outcome was binary. To explore potential interaction effects, interaction terms between ACS and each subgroup variable were included in the logistic regression to identify significant associations.

RCS

Due to the limitations of the overlapping weighting method, RCS analysis in this study was conducted only on the cohort after PSM. RCS were utilized to examine whether there is a nonlinear relationship between gestational age at delivery and the incidence of NRDS, NRF, neonatal pneumonia, and neonatal respiratory composite outcom among all preterm pregnancies, and to observe the trend in odds ratios (ORs). Four knots were set with the reference value for gestational age at delivery being the median of 34.7 weeks.

Sensitivity analyses

Given that the absolute standardized mean differences for nearly all covariates were less than 10%, 1:1 PSM was employed as an alternative to overlap weighting [12]. Logistic regression was then used to perform separate analyses of the association between neonatal outcomes and ACS administration, as well as subgroup analyses. The above analysis treated each twin as an individual.

Results

Baseline characteristics and clinical features of the ACS and non-ACS groups

After screening for exclusion criteria among 2,153 eligible twin pregnancies, 156 were excluded. Consequently, the study included 1,997 twin pregnancy cases, of which 734 (36.8%) received ACS treatment before 36 weeks and 6 days of gestation, and 1,264 (63.2%) did not receive ACS treatment. Baseline characteristic analysis revealed that DCDA pregnancies (p = 0.047), ART (p < 0.001), nulliparity (p < 0.001), placenta previa (p < 0.001), placenta accreta spectrum (p = 0.029), and the occurrence of ICP (p < 0.001) were associated with ACS treatment (Table 1). Notably, there was a significantly lower proportion of scarred uterus in the ACS-treated group, and the gestational age was significantly shorter compared to the non-ACS group (34.50 [33.50, 35.70] vs. 34.80 [33.70, 36.00]). There were no significant statistical differences between the two groups in terms of maternal age, pre-gestation BMI, primigravida, family history of hypertension, family history of diabetes, family history of cancer, GDM, PE, hypothyroidism, anaemia, thrombopenia, and other baseline characteristics. Moreover, there were no significant differences in demographic baselines and clinical features between the two groups in cases of early preterm birth. However, in cases of late preterm birth, the ACS-treated group had significantly higher proportions of ART, nulliparity, uterine myoma, placenta previa, placenta accreta spectrum, and risks of ICP and PE compared to the non-ACS group, while the proportion with a scarred uterus and the gestational age at delivery were significantly lower. The characteristics and clinical features of the mother, calculated by treating each twin as two separate individuals, can be found in Supplementary Table 3.

Table 1.

Baseline characteristics of the study cohorts

Characteristica	All-PTB				Early-PTB				Late-PTB
	No-ACS group (n = 1263)	ACS group (n = 734)	p-value	ASMD	No-ACS group (n = 167)	ACS group (n = 304)	p-value	ASMD	No-ACS group (n = 1096)	ACS group (n = 430)	p-value	ASMD
Maternal age, Median (IQR)	31.00[28.00,33.00]	31.00[28.00,33.00]	0.743	0.002	31.00[28.00,33.00]	31.00[28.00,33.00]	0.989	0.016	31.00[28.00,33.00]	31.00[28.00,34.00]	0.426	0.012
Pre-gestation BMI, Median (IQR)	21.23[19.57,23.42]	21.48[19.53,23.50]	0.478	0.197	21.50[19.71,23.88]	21.640[19.910,23.620]	0.890	0.030	21.23[19.56,23.24]	21.33[19.530,23.42]	0.699	0.025
DCDA, n(%)	978(77.44)	596(81.20)	0.047*	0.093	132(79.04)	245(80.59)	0.687	0.039	846(77.19)	351(81.63)	0.058	0.058
ART, n(%)	838(66.35)	550(74.93)	< 0.001*	0.189	115(68.86)	224(73.68)	0.265	0.106	723(65.97)	326(75.814)	< 0.001*	0.088
Gestational age, median(IQR)	34.80[33.70,36.00]	34.50[33.50,35.70]	0.008*	0.709	32.50[30.80,33.30]	32.50[31.00,33.30]	0.820	0.023	36.00[35.30,36.50]	35.30[34.50,36.00]	< 0.001*	0.014
Primigravida, n(%)	573(45.37)	342(46.59)	0.596	0.025	72(43.11)	151(49.67)	0.173	0.131	501(45.71)	191(44.42)	0.648	0.099
Nulliparity, n(%)	982(77.75)	622(84.74)	< 0.001*	0.180	137(82.04)	262(86.18)	0.231	0.113	845(77.10)	360(83.72)	0.004*	0.104
Family history of hypertension, n(%)	165(13.06)	109(14.85)	0.263	0.052	24(14.37)	40(13.16)	0.713	0.035	141(12.87)	69(16.05)	0.105	0.100
Family history of diabetes, n(%)	83(6.57)	36(4.91)	0.129	0.072	12(7.19)	12(3.95)	0.126	0.141	71(6.48)	24(5.58)	0.514	0.105
Family history of cancer, n(%)	34(2.69)	18(2.45)	0.746	0.015	7(4.19)	5(1.65)	0.093	0.151	27(2.464)	13(3.02)	0.538	0.101
Scarred uterus, n(%)	159(12.59)	66(8.99)	0.014*	0.116	14(8.38)	25(8.22)	0.952	0.006	145(13.230)	41(9.54)	0.047*	0.089
Uterine myoma, n(%)	46(3.64)	34(4.63)	0.277	0.050	10(5.99)	9(2.96)	0.110	0.146	36(3.285)	25(5.81)	0.023*	0.046
Placenta previa, n(%)	39(3.09)	53(7.22)	< 0.001*	0.188	9(5.39)	20(6.58)	0.607	0.050	30(2.737)	33(7.67)	< 0.001*	0.023
Placenta accreta spectrum, n(%)	175(13.86)	157(21.39)	< 0.001*	0.199	32(19.16)	62(20.40)	0.749	0.031	143(13.047)	95(22.09)	< 0.001*	0.063
GDM, n(%)	365(28.90)	240(32.70)	0.075	0.082	55(32.93)	97(31.91)	0.820	0.022	310(28.29)	143(33.26)	0.056	0.011
ICP, n(%)	229(18.13)	185(25.20)	< 0.001*	0.172	31(18.56)	60(19.74)	0.758	0.030	198(18.07)	125(29.07)	< 0.001*	0.006
PE, n(%)	196(15.52)	130(17.71)	0.201	0.059	14(8.38)	27(8.88)	0.854	0.018	182(16.61)	103(23.95)	< 0.001*	0.065
Gestational hypothyroidism, n(%)	115(9.11)	68(9.26)	0.905	0.006	15(8.98)	27(8.88)	0.971	0.004	100(9.12)	41(9.54)	0.803	0.065
Gestational anaemia, n(%)	346(27.40)	204(27.79)	0.848	0.009	55(32.93)	98(32.24)	0.877	0.015	291(26.55)	106(24.65)	0.447	0.090
Gestational thrombopenia, n(%)	52(4.12)	28(3.82)	0.740	0.015	7(4.19)	7(2.30)	0.248	0.106	45(4.11)	21(4.88)	0.502	< 0.001

a: Definitions of characteristics measures listed in Supplementary Table 1

Abbreviation: PTB, Preterm birth; ACS, Antenatal corticosteroid; DCDA, Dichorionic-Diamniotic; ART, Assisted reproductive technology; GDM, Gestational Diabetes Mellitus; ICP, Intrahepatic Cholestasis of Pregnancy; PE, Preeclampsia; ASMD, Absolute standardized mean differences. * p < 0.05

Comparison of neonatal outcomes between the ACS and non-ACS groups

The SMDs after overlap weighting were less than 0.1 and stable in both the overall preterm population and the subgroups of early and late preterm births, indicating that the matching was good (Supplementary Fig. 1–3).

In the total population of preterm births after overlap weighting adjustment, univariate logistic regression analysis found no significant difference in the risk of neonatal respiratory composite outcomes between the ACS treatment group and the untreated group (weighted OR 0.84 [95% CI, 0.70–1.01]). Similarly, there was no significant difference in the risk of NRF (weighted OR 0.97 [95% CI, 0.77–1.21]). However, ACS treatment significantly reduced the risk of NRDS and neonatal pneumonia (weighted OR for NRDS 0.63 [95% CI, 0.49–0.81]; weighted OR for neonatal pneumonia 0.67 [95% CI, 0.52–0.86]). Additionally, ACS treatment was associated with a lower risk of LBW, SGA infants, neonatal purpura, and neonatal hypoproteinemia (weighted OR for LBW 0.66 [95% CI, 0.56–0.77]; weighted OR for SGA 0.61 [95% CI, 0.40–0.93]; weighted OR for neonatal purpura 0.31 [95% CI, 0.16–0.59]; weighted OR for neonatal hypoproteinemia 0.43 [95% CI, 0.27–0.69]). Even after weighting, ACS treatment was associated with an increased risk of hypoglycemia and hyperbilirubinemia in neonates born to mothers who received the treatment. (weighted OR for neonatal hypoglycemia 1.58 [95% CI, 1.14–2.19]; weighted OR for neonatal hyperbilirubinemia 1.60 [95% CI, 1.36–1.89]). Other neonatal complications such as neonatal enteritis, polyhydramnios, and oligohydramnios showed no significant difference in risk (Table 2).

Table 2.

Association of ACS administration with neonatal outcomes after overlap weighting

Outcomesa	All-PTB				Early-PTB				Late-PTB
	No-ACS group, No. (%)	ACS group, No. (%)	OR (95% CI)	p-value	No-ACS group, No. (%)	ACS group, No. (%)	OR (95% CI)	p-value	No-ACS group, No. (%)	ACS group, No. (%)	OR (95% CI)	p-value
Primary outcomes (NRDS)	11.6	7.6	0.63[0.49,0.81]	< 0.001*	33.1	26.9	0.74[0.55,1.00]	0.052	3.3	2.8	0.84[0.51,1.37]	0.480
Secondary outcomes
Neonatal respiratory outcome
NRF	11.5	11.2	0.97[0.77,1.21]	0.765	30.4	30.9	1.03[0.76,1.38]	0.864	4.8	4.7	0.98[0.66,1.45]	0.903
Neonatal pneumonia	10.5	7.3	0.67[0.52,0.86]	0.002*	29.2	17.6	0.52[0.37,0.71]	< 0.001*	4.6	2.8	0.61[0.39,0.96]	0.034*
Neonatal respiratory composite outcomeb	19.5	16.9	0.84[0.70,1.01]	0.066	51.9	46.7	0.81[0.62,1.07]	0.137	8.3	7.0	0.83[0.61,1.15]	0.261
Neonatal digestive outcome
Neonatal enteritis	1.7	1.7	0.99[0.59,1.66]	0.973	1.9	3.9	2.06[0.87,4.91]	0.102	1.8	0.8	0.41[0.19,0.91]	0.028*
Lower gastrointestinal hemorrhage	5.6	3.8	0.67[0.48,0.95]	0.023*	8.5	5.3	0.60[0.36,1.02]	0.059	5.1	2.5	0.47[0.29,0.77]	0.003*
Neonatal digestive complex outcome	7.1	5.5	0.76[0.57,1.02]	0.067	10.1	9.2	0.90[0.57,1.42]	0.655	6.7	3.2	0.46[0.30,0.71]	< 0.001*
Other outcomes
Neonatal Infection	3.2	1.8	0.56[0.35,0.89]	0.015*	7.2	4.3	0.57[0.32,1.02]	0.058	2.0	0.9	0.45[0.21,0.99]	0.046*
LBW	32.6	24.1	0.66[0.56,0.77]	< 0.001*	54.9	36.4	0.47[0.36,0.62]	< 0.001*	28.7	17.2	0.52[0.42,0.64]	< 0.001*
SGA	3.8	2.3	0.61[0.40,0.93]	0.021*	4.9	2.8	0.55[0.27,1.10]	0.092	3.6	2.1	0.59[0.34,1.03]	0.061
Polyhydramnios	1.4	1.5	1.05[0.59,1.85]	0.876	0.8	1.6	2.08[0.55,7.91]	0.280	1.6	1.4	0.87[0.43,1.75]	0.697
Oligohydramnios	3.1	4.4	1.43[0.99,2.07]	0.056	6.0	7.2	1.22[0.70,2.12]	0.477	2.4	2.8	1.17[0.70,1.98]	0.548
Neonatal purpura	2.3	0.7	0.31[0.16,0.59]	< 0.001*	5.6	2.4	0.41[0.20,0.86]	0.019*	1.3	0.2	0.13[0.03,0.60]	0.009*
Neonatal hypoglycemia	3.7	5.8	1.58[1.14,2.19]	0.007*	5.0	5.4	1.09[0.59,2.02]	0.781	3.6	5.6	1.60[1.08,2.39]	0.020*
Neonatal hyperbilirubin	21.3	30.2	1.60[1.36,1.89]	< 0.001*	42.7	51.5	1.43[1.09,1.88]	0.011*	15.7	19.2	1.28[1.02,1.60]	0.031*
Neonatal hypoproteinemia	3.5	1.5	0.43[0.27,0.69]	< 0.001*	7.8	4.2	0.52[0.29,0.91]	0.022*	2.6	0.6	0.22[0.09,0.57]	0.002*
Neonatal hyperlacticemia	4.2	2.8	0.66[0.44,0.97]	0.036*	7.4	3.5	0.45[0.24,0.83]	0.010*	3.5	2.5	0.69[0.41,1.15]	0.151

a: Definitions of outcome measures listed in Supplementary Table 2

b: Neonatal respiratory composite outcome defined as at least one of the following occurrences in NRDS, NRF or neonatal pneumonia

c: Neonatal digestive composite outcome defined as at least one of the following occurrences in neonatal enteritis or lower gastrointestinal hemorrhage

Abbreviation: NRDS, neonatal respiratory distress syndrome; NRF, Neonatal respiratory failure; LBW, Low birth weight infant; SGA, Small for gestational age; NICU admission, Neonatal Intensive Care Unit admission. * p < 0.05

Further subgroup analyses for early and late preterm births revealed that there were no significant differences in primary outcome NRDS between the ACS-treated and untreated groups, whether in early or late preterm births (early preterm p = 0.052; late preterm p = 0.480). Similarly, there was no significant difference in the risk of NRF and neonatal respiratory composite outcomes (NRF early preterm p = 0.864, late preterm p = 0.903; Neonatal respiratory composite outcomes early preterm p = 0.137, late preterm p = 0.261). However, ACS treatment was associated with a reduced risk of neonatal pneumonia in both early and late preterm births. (weighted OR for early preterm 0.47 [95% CI, 0.36–0.62]; weighted OR for late preterm 0.61 [95% CI, 0.39–0.96]). Moreover, ACS treatment was associated with a reduced risk of LBW and neonatal hypoproteinemia in both early and late preterm births. (weighted OR for early preterm LBW 0.47 [95% CI, 0.36–0.62]; weighted OR for late preterm LBW 0.52 [95% CI, 0.42–0.64]; weighted OR for early preterm neonatal hypoproteinemia 0.52 [95% CI, 0.29–0.91]; weighted OR for late preterm neonatal hypoproteinemia 0.22 [95% CI, 0.09–0.57]).

Regarding outcomes related to the digestive system, ACS treatment in late preterm births was associated with a reduced risk of neonatal enteritis and lower gastrointestinal bleeding. (weighted OR for neonatal enteritis 0.41 [95% CI, 0.19–0.91]; weighted OR for lower gastrointestinal bleeding 0.47 [95% CI, 0.29–0.77]), and was also associated with a lower risk of neonatal infection (weighted OR for neonatal infection 0.45 [95% CI, 0.21–0.99]). In early preterm births, however, ACS treatment was not associated with a significant reduction in the risks of these conditions. Additionally, for neonatal hyperlacticemia, ACS treatment was associated with a reduced risk in early preterm births (weighted OR 0.69 [95% CI, 0.41–1.15]), although this effect was not significant in late preterm births. Concurrently, the risks of neonatal hypoglycemia and hyperbilirubinemia were associated with an increased likelihood after ACS use.

Subgroup and interaction analyses

To further investigate the mitigating effects of ACS on the occurrence of primary outcome NRDS and neonatal respiratory composite outcome across different periods of preterm birth, subgroup and interaction analyses were conducted based on chorionicity (whether DCDA or not), use of ART, primiparity, and the presence of GDM, ICP, and PE.

After applying overlap weighting, subgroup and interaction analyses revealed significant findings. In early preterm births, among twin pregnancies complicated by GDM, ACS treatment not only failed to reduce risks but also significantly increased the incidence of NRDS and the neonatal respiratory composite outcome. Specifically, the incidence of NRDS was 1.89 (95% CI, 1.08–3.30) and the neonatal respiratory composite outcome was 1.68 (95% CI, 1.03–2.76), both of which showed significant interaction with GDM (p for interaction < 0.05). In contrast, for twin pregnancies complicated by PE, ACS treatment significantly increased the incidence of NRDS in both the overall preterm population and the early preterm subgroup. The relative risks for NRDS were 2.19 (95% CI, 1.09–4.37) for all preterm births and 4.19 (95% CI, 1.21–14.48) for early preterm births, with interaction test p-values of 0.013 and 0.018, respectively(Table 3). For other maternal characteristics, ACS was effective in reducing the neonatal respiratory composite outcome only in twin pregnancies of the DCDA type and those without ART, with relative risks of 0.80 (95% CI, 0.65–0.99) for DCDA and 0.57 (95% CI, 0.34–0.95) for no ART (Supplementary Table 4). No significant effects were observed for other maternal subgroups. These findings suggest that the efficacy of ACS treatment varies in a complex manner depending on the stage of preterm birth and the presence of maternal complications. Specifically, early preterm births complicated by GDM and PE may serve as risk factors for adverse neonatal outcomes associated with ACS treatment.

Table 3.

Association of antenatal corticosteroids with neonatal respiratory distress syndrome in the prespecified subgroups after overlap weighting

Variables	All-PTB			Early-PTB			Late-PTB
	OR(95%CI)	P-value	P for interaction	OR(95%CI)	P-value	P for interaction	OR(95%CI)	P-value	P for interaction
DCDA			0.764			0.273			0.869
No	1.03 [0.63, 1.68]	0.899		0.75 [0.40, 1.42]	0.375		1.05 [0.42, 2.60]	0.923
Yes	0.95 [0.73, 1.23]	0.688		1.12 [0.80, 1.57]	0.508		0.96 [0.62, 1.49]	0.857
ART			0.988			0.188			0.610
No	0.97 [0.65, 1.45]	0.875		0.77 [0.45, 1.30]	0.319		0.83 [0.40, 1.72]	0.620
Yes	0.96 [0.73, 1.27]	0.799		1.17 [0.82, 1.68]	0.382		1.04 [0.65, 1.68]	0.860
Primipara			0.945			0.084			0.705
No	0.98 [0.57, 1.70]	0.951		0.55 [0.26, 1.18]	0.127		1.13 [0.48, 2.62]	0.783
Yes	0.96 [0.75, 1.24]	0.765		1.14 [0.83, 1.57]	0.425		0.94 [0.60, 1.47]	0.774
GDM			0.480			0.008*			0.288
No	0.92 [0.70, 1.21]	0.534		0.78 [0.55, 1.11]	0.166		1.14 [0.70, 1.87]	0.596
Yes	1.09 [0.73, 1.65]	0.665		1.89 [1.08, 3.30]	0.025*		0.73 [0.37, 1.42]	0.352
ICP			0.289			0.433			0.780
No	0.91 [0.71, 1.17]	0.477		0.98 [0.71, 1.35]	0.879		0.95 [0.60, 1.48]	0.811
Yes	1.26 [0.73, 2.18]	0.401		1.36 [0.63, 2.95]	0.433		1.09 [0.46, 2.54]	0.850
PE			0.013			0.018*			0.255
No	0.87 [0.68, 1.11]	0.252		0.92 [0.67, 1.25]	0.576		0.87 [0.55, 1.36]	0.531
Yes	2.19 [1.09, 4.37]	0.027*		4.19 [1.21, 14.48]	0.024*		1.55 [0.63, 3.77]	0.338

Abbreviation: PTB, Preterm birth; ACS, Antenatal corticosteroid; DCDA, Dichorionic-Diamniotic; ART, Assisted reproductive technology; GDM, Gestational Diabetes Mellitus; ICP, Intrahepatic Cholestasis of Pregnancy; PE, Preeclampsia. * p < 0.05

Impact of ACS treatment on neonatal morbidity across different gestational ages

To further visualize the impact of ACS on the occurrence of neonatal respiratory system disease outcomes at different gestational ages, we conducted an in-depth analysis on the total preterm birth population using the restricted cubic splines method. This was aimed at revealing the influence of ACS on neonatal respiratory system disease outcomes across various gestational ages. The analysis clearly indicated that gestational age is an important factor influencing the risk of the primary outcome, NRDS, regardless of ACS treatment (p < 0.001), and its impact on NRDS showed a significant non-linear relationship (p for nonlinearity < 0.001). Although there was overlap in the confidence intervals for the risk trends of NRDS and other respiratory diseases, including neonatal respiratory composite outcomes, between neonates with and without ACS treatment, preventing statistically significant differences, the trends suggest that ACS treatment may help reduce the risks of NRDS, neonatal pneumonia, and neonatal respiratory composite outcomes across different gestational ages. This finding should be interpreted with caution and further explored in future research to better understand the effects of ACS treatment on neonatal respiratory system disease outcomes at various gestational weeks (Fig. 2).

Fig. 2.

The association between preterm twin pregnancy delivery and the occurrence of a NRDS, as well as the odds ratios for various respiratory outcomes including NRF, neonatal pneumonia and neonatal respiratory composite outcome, was determined using restricted cubic spline fitting. (a) NRDS. (b) NRF. (c) Neonatal pneumonia. (d) Neonatal respiratory composite outcome. The reference level for the odds ratios (ORs) is the median gestational age at delivery, which is 34.7 weeks. The reference line is Y = 1

Sensitivity analysis

In this study, we conducted comprehensive sensitivity analyses to verify the robustness of our findings. Initially, we utilized a 1:1 PSM as an alternative to OW (Supplementary Table 5). Consistent with the main analysis, we did not observe positive results regarding the NRDS for the primary outcome (All-PTB, p = 0.182; Early-PTB, p = 0.337; Late-PTB, p = 0.168) as shown in Supplementary Table 6. Furthermore, the subgroup analyses after PSM, largely in agreement with the main analysis, indicated that the presence of GDM and PE significantly affects the efficacy of ACS on the NRDS and neonatal respiratory composite outcomes, further confirming the robustness of our study results (Supplementary Tables 7 and 8).

Discussion

Previous clinical practice guidelines have recommended that ACS treatment for twin and higher-order multiple pregnancies should be consistent with that for singleton pregnancies. This recommendation is based on evidence from proven therapeutic effects in singleton trials, including several cohort studies that have demonstrated ACS significantly reduces important neonatal outcomes such as mortality and surfactant use [5, 23]. However, the results from meta-analyses of multiple gestation trials, as well as non-randomized studies, remain uncertain, often reporting a protective effect but with less consistent findings [24, 25]. For instance, the latest Green-top guidelines from the UK and a non-randomized study on twins reported no effect or harmful impacts, hence the lower grading of the recommendation [26]. Addressing this issue, our study aimed to provide more robust evidence to support the recommendations for twins by designing a study based on a large twin pregnancy cohort and adjusting for confounders through various matching methods. This approach enhances precision relative to trial data and reduces the risk of confounding compared to traditional non-randomized study data. While we excluded monochorionic complications, such as TTTS, to focus on a more generalizable twin population, we acknowledge that the results may not fully apply to high-risk monochorionic twins.

Our study found that ACS significantly reduced the risk of NRDS in all preterm twin pregnancies. This finding is consistent with the results of the 2020 Cochrane systematic review (RR = 0.71, 95% CI: 0.65–0.78) [4] and the conclusions of a 2022 meta-analysis of non-randomized studies on twins (OR = 0.70, 95% CI: 0.57–0.86) [6]. Additionally, our study revealed that ACS significantly reduced the risk of neonatal pneumonia. Preterm birth is the most critical risk factor for neonatal infections, and neonatal pneumonia is the second most common preterm-related infection, closely associated with respiratory diseases [27]. Previous studies have indicated that pneumonia is not only included as a key outcome indicator in surfactant therapy guidelines [28] but also that corticosteroid therapy may effectively prevent neonatal pneumonia, as observed in studies consistent with our findings [29, 30]. Therefore, these results provide strong evidence supporting the benefits of ACS in twin pregnancies at risk of preterm birth and further reinforce the recommendation for this therapy in current clinical practice guidelines.

Previous studies have recommended a single course of corticosteroids for women at risk of preterm birth from 24 weeks 0 days to 33 weeks 6 days, regardless of the number of fetuses [24, 25]. However, the evidence supporting this recommendation is inferred from singleton randomized controlled trials, although some studies on twin pregnancies also suggest that anti-inflammatory corticosteroids can have the same or comparable effect in preventing RDS in preterm twins [10, 31, 32]. Nevertheless, this study found that ACS did not significantly reduce the occurrence of NRDS and neonatal respiratory composite outcomes in early preterm pregnancies, only having a significant effect in neonatal pneumonia. This may be due to pharmacokinetic differences between twins and singletons, such as shorter half-life and higher clearance of ACS in twins. Additionally, physiological changes in maternal blood volume during twin pregnancies may affect drug distribution, thus impacting the effectiveness of ACS intended to promote fetal lung maturity [33, 34]. This suggests that the effectiveness of ACS in preventing RDS in early preterm births should be evaluated with caution. Moreover, the lack of statistical significance observed in the primary outcome of NRDS, as well as in NRF and neonatal respiratory composite outcomes, may be attributable to β error caused by the limited sample size of women with early preterm birth [35]. However, this study included 471 twin pregnancies and treated each neonate from the same mother as an independent individual in the analysis, thereby enhancing the reliability of the findings. Future studies could further increase the sample size within this subgroup to improve statistical power and validate the conclusions of the present study. In addition, the study also found that ACS significantly reduced the risk of LBW, neonatal hypoproteinemia, and neonatal hyperlacticemia in early preterm neonates, indicating that ACS may regulate metabolism by promoting balanced metabolism of glucose, fat, and protein, and reducing protein loss, which helps fetal growth and thus reduces the risk of low birth weight in preterm infants.

While there is considerable evidence supporting the use of ACS in late preterm singleton pregnancies [36–38]the evidence for twin pregnancies is limited [11, 13]. Most studies to date have found that antenatal corticosteroid treatment for women at risk of late preterm birth does not reduce the incidence of NRDS [11, 13] and some have found that higher rates of antenatal corticosteroid administration reduce the baseline risk of respiratory diseases in twins. Therefore, it is necessary to conduct high-quality research to explore the therapeutic effect of late preterm ACS on respiratory diseases by adequately adjusting for confounders. A recent study from China on twin pregnancies at risk of late preterm birth using overlap weighting analysis showed that ACS did not significantly reduce the risk of NRDS and neonatal respiratory composite outcomes, but may increase the risk of neonatal hypoglycemia [12]. This is consistent with the conclusions of this study. Interestingly, this study also found that ACS showed significant effects on neonatal pneumonia, which may be because ACS accelerates the maturation of the fetal lungs, particularly by promoting the production of pulmonary surfactants, which helps maintain the openness of alveoli and reduce breathing difficulties. The anti-inflammatory actions of ACS may also help reduce lung inflammation, thereby lowering the risk of neonatal pneumonia due to immature lungs [39]. Additionally, ACS not only increases the risk of neonatal hypoglycemia but also increases the risk of neonatal hyperbilirubinemia, consistent with an Israeli study [40]. Our research also extends the assessment of late preterm neonatal outcomes, finding that ACS associated with a significant reduction in the risks of neonatal hypoproteinemia, neonatal infection, and significant decreases in digestive outcomes such as neonatal enteritis and lower gastrointestinal hemorrhage [41]. This may be because cortisol helps to mature the fetal intestinal barrier in advance, enhancing immune function, and thereby reducing the risk of neonatal enteritis and other digestive diseases [42]. Additionally, this study found that ACS was significantly associated with a reduction in the risk of LBW, regardless of early or late preterm birth, consistent with previous singleton studies. Meanwhile, research suggests that ACS may affect birth weight, particularly with ≥ 4 courses [43]. Murphy et al. reported that repeated ACS significantly reduced birth weight, length, and head circumference, while Noel P. French’s study further confirmed that an increased number of ACS courses led to a 9% reduction in birth weight (P = 0.014) and a 4% reduction in head circumference (P = 0.0024) [44, 45], suggesting a cumulative effect on fetal growth. However, as this study did not record dosage or repeated courses, future research should further explore the impact of repeated ACS on LBW, particularly in twin pregnancies, to optimize clinical management. It should be noted that while our study has shown associations between ACS treatment and various neonatal outcomes, the causal conclusions drawn are limited to the primary outcome for which the propensity score was calculated. For these secondary outcome, although trends were observed, the study design precludes firm causal inferences [46].

In previous studies of late preterm births, subgroup analyses of gestational age at birth, year of birth, chorionicity, at least one infant with SGA, discordant twin growth, and infant gender did not show significant interactions between ACS and these factors on neonatal respiratory outcomes [12]. In this study, although there was no significant interaction between chorionicity and ACS in late preterm pregnancies, whether the outcome was NRDS or neonatal respiratory composite outcomes., ACS significantly reduced the incidence of neonatal respiratory composite outcomes in twin pregnancies with DCDA chorionicity, but not in non-DCDA. Additionally, the study found that the effectiveness of ACS treatment in early preterm neonates for NRDS and neonatal respiratory composite outcomes was significantly related to the presence of GDM. GDM increases insulin resistance, leading to elevated maternal blood glucose, which may affect fetal lung development and increase the risk of neonatal respiratory complications [47, 48]. Although pregnant women with GDM have a higher risk of preterm birth and typically require ACS treatment, corticosteroids themselves have a hyperglycemic effect, which, if maternal blood glucose is poorly controlled, may worsen respiratory system diseases in preterm infants. Previous studies have shown that intrauterine glucose infusion can alter corticosteroid signaling, inhibiting the expression of surfactant protein mRNA in fetal lungs [49]. Additionally, GDM pregnant women exhibit significantly elevated blood glucose after receiving betamethasone treatment [50]which may weaken the effect of ACS treatment, especially in early preterm births, thereby increasing the risk of NRDS [51]. Therefore, in a hyperglycemic environment, the benefits of ACS treatment may be counteracted and even exacerbate certain respiratory complications. The findings of this study suggest that the hyperglycemic state in GDM pregnancies may be more pronounced in twin pregnancies with early preterm birth, affecting the efficacy of ACS treatment. Therefore, optimizing maternal blood glucose control, especially when using ACS, may be key to improving treatment outcomes and reducing the risk of neonatal respiratory complications. Finally, this study found that ACS treatment significantly increased the incidence of NRDS in both the overall preterm and early preterm subgroups of twin pregnancies complicated by PE. PE is commonly associated with placental dysfunction and systemic inflammation, which may reduce fetal responsiveness to ACS and interfere with fetal lung development, thereby diminishing the effect of ACS in promoting lung maturation [52, 53]. Although some studies suggest that ACS is beneficial for preterm infants born to mothers with severe PE, including significantly reducing the incidence of NRDS and other complications, these studies mainly focus on singleton pregnancies, severe PE cases, and have limited sample sizes, which may affect the generalizability of the results [49, 54]. Additionally, other studies have indicated that corticosteroid treatment in severe PE pregnancies may result in lower neonatal birth weight, a longer interval between admission and delivery, and potentially worse outcomes such as stillbirth and neonatal death [50]. These results suggest that PE may affect the efficacy of ACS treatment in twin pregnancies, and thus, the effectiveness of ACS may be diminished in pregnancies complicated by PE. Future research should further explore the specific impact of PE on ACS treatment outcomes in early preterm births, particularly in twin pregnancies, to provide more precise clinical treatment strategies.

Strengths and limitations

To minimize potential confounding factors in antenatal corticosteroid treatment in twin pregnancies, we implemented a series of rigorous strategies. Specifically, we employed propensity score overlap weighting matching, a method that more effectively eliminates potential confounders compared to traditional multivariable logistic regression, achieving the closest effect to a randomized clinical trial in a retrospective study [17]. Given the lack of randomized clinical trials on the impact of antenatal corticosteroid treatment in twin pregnancies, we believe this carefully designed observational study can offer valuable insights for clinical decision-making. Furthermore, the robustness of our findings, confirmed through multiple subgroup and sensitivity analyses, enhances the credibility and reliability of the results.

However, Firstly, causal interpretations require circumspection, as robust causal linkages were only substantiated for primary outcomes (NRDS) and related respiratory outcomes. Although trends were observed in secondary outcomes (e.g., gastrointestinal outcomes), they did not support definitive causal conclusions. Due to the retrospective observational design, despite efforts to balance group differences using propensity score matching, unmeasured confounders could not be fully eliminated, and the conditions of a randomized controlled trial (RCT) were not fully replicated. Due to the lack of detailed records on respiratory support measures (such as CPAP, HFNC, FiO₂, ECMO, and mechanical ventilation), we were unable to fully assess respiratory-related outcomes. Typically, outcomes are defined based on factors like surfactant use, Bronchopulmonary dysplasia (BPD) incidence, or composite outcomes including neonatal mortality. These aspects could be considered in future analyses of respiratory composite outcomes. Additionally, although hypoxia-induced complications (such as neonatal intraventricular hemorrhage) may be associated with respiratory diseases, this study primarily focused on directly and clinically common respiratory outcomes, such as NRDS. Therefore, the impact of these complications was not fully considered, and future research may further explore this aspect [55, 56]. Due to the lack of detailed data on the timing of ACS administration, we were unable to accurately distinguish whether ACS was given during the early or late preterm period. This limitation complicates the definition of “time zero,” as we cannot pinpoint the exact timing of ACS relative to delivery [57, 58]. Including all twin pregnancies in the analysis without this critical information may misclassify full-term pregnancies into the Non-ACS group, introducing confounding bias. While we restricted our analysis to preterm pregnancies to minimize this issue, potential confounding remains, especially for those who received ACS before 37 weeks but delivered at term. Our reliance on real-world clinical data also limits the accuracy of variables like ACS timing, dose, and frequency, further complicating causal inference. Factors such as chorioamnionitis and intrauterine fetal death, which can emerge later in pregnancy, add additional confounding challenges [12, 15]. Although we excluded these cases to improve homogeneity, this still weakens the causal interpretation Future studies should adopt a prospective design with accurate recording of ACS timing and delivery specifics. This would allow for a more precise definition of “time zero” and reduce confounding by enabling stratified analyses based on ACS timing. Such studies would better simulate the ideal conditions of a target trial and improve the validity of causal conclusions [59–61]. Lastly, we did not collect detailed information on key factors such as the specific types of corticosteroids used, the frequency and completeness of dosing, and the timing of administration relative to delivery. These variables are critical, as incomplete or inconsistent corticosteroid courses, along with variations in dosing schedules, can significantly influence neonatal outcomes. The absence of this data introduces potential bias and limits the ability to accurately assess the true impact of corticosteroid treatment on the study’s results. In addition, we did not apply correction for multiple testing, as all analyses were exploratory and consistent with previous research methods. This may increase the risk of finding low p-values by chance, and the results should be interpreted with caution [22]. These limitations suggest that in future research, these variables need to be considered more meticulously and comprehensively to further validate our findings.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Wei-Zhen Tang, Conceptualization, Methodology, Software, Data Curation, Original Draft Preparation, Visualization; Wei-Ze Xu, Software, Investigation, Visualization; Qin-Yu Cai, Conceptualization, Methodology, Validation, Investigation, Data Curation, Original Draft Preparation, Supervision, Project Administration; Kang-Jin Huang, Investigation; Hong-Yu Xu, Investigation; Jia-Zheng Li, Investigation; Bo-Yuan Deng, Investigation; Hao-Wen Chen, Investigation; Li Wen, Investigation; Lan Wang, Conceptualization, Data Curation, Writing-Review and Editing, Supervision, Project Administration; Tai-Hang Liu, Conceptualization, Methodology, Investigation, Data Curation, Original Draft Preparation, Writing-Review and Editing, Supervision, Project Administration.

Funding

This work was funded by the National Key Research and Development Program of China (No. 2023YFC2705900), the Natural Science Foundation of Chongqing (No. CSTB2024NSCQ-MSX0706 and CSTB2023NSCQ-MSX0384), and the National Training Program of Innovation and Entrepreneurship for Undergraduates (No. 202410631039X).

Data availability

The data underlying this article will be provided by the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Women and Children’s Hospital of Chongqing Medical University (ID: 2022-011-01). To protect patient privacy, all personal identifying information was removed from the cases, and all data obtained were kept anonymous.

Consent for publication

Not applicable.

Conflict of interest

The authors declared that they have no conflict of interest.