Does the timing of fetal reduction from dichorionic twin pregnancies to singleton pregnancies influence premature birth risk? A retrospective cohort study.

The flowchart of the cohort was show in Fig.  1 . In total, 463 DC twin pregnancies were included in the study. Table 1 displays the demographic and clinical characteristics of the study cohort. Most women with DC twin pregnancies concepted from IVF/ICSI (81.4%), while 8.4% conceived through spontaneous conception, 6.7% through ovulation induction and 3.5% concepted from IUI. The mean GA at reduction was 10.3 ± 3.8 weeks. Among the women, 350 (75.6%) underwent transvaginal FR and 113 (24.4%) underwent transabdominal FR, while 343 patients with FR time  11w. The indications for FR were sonographic anatomical abnormality or chromosomal anomaly (8.2%), poor obstetric history (6.7%), adverse uterine factors (46.9%) and miscellaneous factors (38.2%). The GA distribution of reduction for different indications was show in Fig.  1 . Selective terminations were mainly performed in the second trimester, while elective FRs were mostly performed at an earlier GA. Table 1 Demographic and clinical characteristics of 463 women with dichorionic twin pregnancies underwent fetal reduction Characteristic NO. (%) or Mean ± SD Maternal characteristic  Maternal age at conception (years) 32.0 ± 3.8  Maternal Height (cm) 158.4 ± 5.6  Maternal pre-conception weight (kg) 54.6 ± 7.7  Maternal pre-conception BMI (kg/m2) 21.7 ± 2.8 Maternal pre-pregnancy weight status (%)  Underweight (BMI < 18.5 kg/m2) 55 (11.9%)  Normal weight (18.5 ≤ BMI < 24 kg/m2) 310 (67.0%)  Overweight (24 ≤ BMI < 28 kg/m2) 89 (19.2%)  Obesity (BMI ≥ 28 kg/m2) 9 (1.9%)  Nulliparous 256 (55.3%) Conception mode  Spontaneous 39 (8.4%)  Ovulation induction 31 (6.7%)  IUI 16 (3.5%)  IVF/ICSI 377 (81.4%) Fetal reduction parameters  Gestational age at reduction (weeks) 10.3 ± 3.8  Transvaginal reduction 350 (75.6%)  Transabdominal reduction 113 (24.4%)  Intra-thoracic potassium chloride 178 (38.4%) Fetal reduction time   11 weeks 120 (25.9%) Indications for fetal reduction  Anatomical abnormality or chromosomal anomaly 38 (8.2%)  Poor obstetric history 31 (6.7%)  Adverse uterine factors 217 (46.9%)  Miscellaneous factors 177 (38.2%) Abbreviations : BMI  Body mass index, IUI  Intrauterine insemination, IVF in-vitro fertilization, ICSI  Intracytoplasmic sperm injection Demographic and clinical characteristics of 463 women with dichorionic twin pregnancies underwent fetal reduction Abbreviations : BMI  Body mass index, IUI  Intrauterine insemination, IVF in-vitro fertilization, ICSI  Intracytoplasmic sperm injection Table S1 indicates GA at delivery (−0.129, P  = 0.007) and birthweight (−0.135, P  = 0.005) were generally negatively correlated with GA at reduction. To intuitively explore the specific impact of GA at reduction on pregnancy outcomes, smooth curve fitting was performed (Fig.  2 ). Figure  2 A shows that pregnancy loss before 24 weeks and GA at reduction were not correlated, with a wide 95% confidence interval (CI) ( P  = 0.428). Figure  2 B-C show that the smooth fitting curves of GA at birth ( P  < 0.001) and birthweight ( P  = 0.025) decrease sharply at later GA, thus a segmented regression model was used to examined the turning point of the GA at reduction (Table  2 ). The same turning point of GA at reduction (17 +4  weeks) was found for both GA at birth (LRT test: P  < 0.001) and birthweight (LRT test: P  = 0.008). It was demonstrated that for every 1-week increase in GA at reduction when FR performed after 17 +4  weeks, GA at birth decreased by −0.51 weeks (−0.75, −0.27) ( P  < 0.001) and birthweight decreased by −100.35 g (−161.98, −38.72) ( P  = 0.002). When FR performed before 17 +4  weeks, GA at birth ( P  = 0.478) and birthweight ( P  = 0.299) were not correlated with GA at reduction. Besides, no turning point around 12–13 weeks was found in the curves, which was the demarcation between transvaginal and transabdominal FR. Fig. 2 Gestational age at reduction and pregnancy outcomes time–response relationship. The associations between the GA at reduction and pregnancy outcomes were shown in smooth curve fitting. *Adjusted for maternal age, height and pre-conception BMI, indications for reduction, conception mode, nulliparous. A  The associations between gestational age at reduction and pregnancy loss < 24 weeks; B  The associations between gestational age at reduction and GA at delivery; C  The associations between gestational age at reduction and birthweight. Abbreviations: GA, gestational age Table 2 Threshold effect analysis of reduction time on gestational age at delivery and birthweight using piece-wise linear regression Reduction time a Adjusted β (95% CI) P value LRT test GA at delivery < 17 +4  weeks −0.03(−0.11, 0.05) 0.478 ≥ 17 +4  weeks −0.51 (−0.75, −0.27)  < 0.001  < 0.001 Birth weight < 17 +4  weeks −10.76 (−31.03, 9.50) 0.299 ≥ 17 +4  weeks −100.35 (−161.98, −38.72) 0.002 0.008 a Adjusted for maternal age, height and pre-conception BMI, indications for reduction, conception mode, nulliparous Abbreviations : LRT test, Log-likelihood ratio test Gestational age at reduction and pregnancy outcomes time–response relationship. The associations between the GA at reduction and pregnancy outcomes were shown in smooth curve fitting. *Adjusted for maternal age, height and pre-conception BMI, indications for reduction, conception mode, nulliparous. A  The associations between gestational age at reduction and pregnancy loss < 24 weeks; B  The associations between gestational age at reduction and GA at delivery; C  The associations between gestational age at reduction and birthweight. Abbreviations: GA, gestational age Threshold effect analysis of reduction time on gestational age at delivery and birthweight using piece-wise linear regression a Adjusted for maternal age, height and pre-conception BMI, indications for reduction, conception mode, nulliparous Abbreviations : LRT test, Log-likelihood ratio test According to the turning point of GA at reduction, patients were divided into two groups as early-reduction group (reduction before 17 +4  weeks, n  = 439) and late-reduction group (reduction after 17 +4  weeks, n  = 24). There was no difference in maternal and clinical characteristics including maternal age at conception, maternal height, maternal pre-conception weight and BMI, nulliparous, or conception mode between two groups (Table S2). In early-reduction group, 79.7% women underwent transvaginal FR, and 20.3% women underwent transabdominal FR, while all women in late-reduction group underwent transabdominal FR ( P  < 0.001). The indications for FR were significantly different between groups (Table S2, P  < 0.001). The most common indications for FR in early-reduction group were adverse uterine factors (49.4%) and miscellaneous factors (39.6%), while the FR in late-reduction group was mostly due to anatomical abnormality or chromosomal abnormalities (83.3%). Unadjusted analyses of pregnancy outcomes in the two groups are shown in Table  3 . There was no significant association observed for pregnancy loss risk before 24 weeks between two groups (4.6% vs 4.2%, P  = 1). Likewise, the rates of pregnancy loss within 2 weeks after FR were also similar between the two groups ( P  = 1). There were two cases of stillbirth in the early-reduction group and none in late-reduction group from 24 weeks ( P  = 1). Similarly, no significant differences were found between two groups with respect to the prevalence of gestational diabetes mellitus (5.5% vs 0.0%, P  = 0.482), hypertensive disease of pregnancy (5.0% vs 0.0%, P  = 0.528) and preterm premature rupture of membranes (1.8% vs 8.3%, P  = 0.090). Table 3 Comparison of pregnancy outcomes of different reduction groups Characteristic Early-FR group Late-FR group P value FR < 17 +4 weeks ( n  = 439) FR ≥ 17 +4 weeks ( n  = 24) Obstetric outcome Pregnancy loss < 24 weeks 20 (4.6%) 1 (4.2%) 1.000 Pregnancy loss with 2 weeks after procedure 3 (0.7%) 0 (0%) 1.000 Fetal death ≥ 24 weeks 2 (0.5%) 0 (0%) 1.000 Live birth 417 (95.0%) 23 (95.8%) 1.000 Gestational diabetes 24 (5.5%) 0 (0%) 0.482 Hypertensive disease of pregnancy 22 (5.0%) 0 (0%) 0.528 Preterm premature rupture of membranes 8 (1.8%) 2 (8.3%) 0.090 Cesarean Delivery 286 (68.6%) 14 (60.9%) 0.439 GA at birth (weeks) 38.1 ± 2.2 36.6 ± 3.3 0.038 Preterm birth < 32 weeks 15 (3.4%) 3 (12.5%) 0.060 Preterm birth < 34 weeks 23 (5.2%) 5 (20.8%) 0.007 Preterm birth < 37 weeks 69 (15.7%) 10 (41.7%) 0.003 Neonatal outcomes  Gender, male (%) 224 (53.7%) 12 (52.2%) 0.885  Birth weight (g) 3075.5 ± 559.4 2761.3 ± 746.1 0.081  Low birth weight (< 2500 g) 52 (11.8%) 8 (33.3%) 0.006  Small for gestational age 35 (8.0%) 3 (12.5%) 0.686 Values presented as n (%) or mean ± SD Comparison of pregnancy outcomes of different reduction groups Values presented as n (%) or mean ± SD Women in late-reduction group were delivered earlier (38.1 ± 2.2 vs 36.6 ± 3.3 weeks, P  = 0.038), with higher rates of preterm birth < 32 weeks (3.4% vs 12.5%, P  = 0.060), < 34 weeks (5.2% vs 20.8%, P  = 0.007) and < 37 weeks (15.7% vs 41.7%, P  = 0.003) than women in the early-reduction group. There was a trend towards lower birthweight in late-reduction group compared to the early-reduction group, but this difference was not significant (3075.5 ± 559.4 vs 2761.3 ± 746.1, P  = 0.081). Compared to the early- reduction group, the late-reduction group had a significantly higher risk of LBW (11.8% vs 33.3%, P  = 0.006), while the risk of SGA was similar between the two groups (8.0% vs 12.5%, P  = 0.665). Logistic regression was performed to compare adverse pregnancy outcomes after adjustment for maternal age, height and pre-conception BMI, indications for reduction, conception mode, nulliparous (Table  4 ). Regression analysis revealed that there remained no significant differences in the risk of pregnancy loss rate (adjusted P  = 0.455) and SGA (adjusted P  = 0.608). The rates of preterm birth < 32 weeks (adjusted OR 4.74 (1.22, 18.41), adjusted P  = 0.025), < 34 weeks (adjusted OR 4.92 (1.64, 14.73), adjusted P  = 0.004) and < 37 weeks (adjusted OR 3.91 (1.64, 9.34), adjusted P  = 0.002) in the late-reduction group, as well as LBW (adjusted OR 2.73 (1.05, 7.08), adjusted P  = 0.040) were significantly higher than those in the early-reduction group, respectively. Table 4 Regression analysis for the association of adverse pregnancy outcomes of different reduction groups Perinatal outcomes Unadjusted ORs (95% CI) Unadjusted P value Adjusted ORs (95% CI) Adjusted P value Pregnancy loss < 24 weeks  Reduction Reduction ≥ 17 +4  weeks 0.91 (0.12, 7.09) 1.000 0.40 (0.04, 4.47) 0.455 Preterm birth < 32 weeks Reduction < 17 +4  weeks Ref Ref  Reduction ≥ 17 +4  weeks 4.04 (1.08, 15.04) 0.060 4.74 (1.22, 18.41) 0.025 Preterm birth < 34 weeks  Reduction < 17 +4  weeks Ref Ref  Reduction ≥ 17 +4  weeks 4.76 (1.63, 13.89) 0.007 4.92 (1.64, 14.73) 0.004 Preterm birth < 37 weeks  Reduction < 17 +4  weeks Ref Ref  Reduction ≥ 17 +4  weeks 3.83 (1.64, 8.97) 0.003 3.91 (1.64, 9.34) 0.002 Low birth weight  Reduction < 17 +4  weeks Ref Ref  Reduction ≥ 17 +4  weeks 3.72 (1.52, 9.12) 0.006 2.73 (1.05, 7.08) 0.040 Small for gestational age  Reduction < 17 +4  weeks Ref Ref  Reduction ≥ 17 +4  weeks 1.65 (0.47, 5.80) 0.686 1.58 (0.27, 9.13) 0.608 Abbreviations : CI confidence interval, ORs odds ratios Adjusted ORs were adjusting for maternal age, height and pre-conception BMI, indications for reduction, conception mode, nulliparous Regression analysis for the association of adverse pregnancy outcomes of different reduction groups Abbreviations : CI confidence interval, ORs odds ratios Adjusted ORs were adjusting for maternal age, height and pre-conception BMI, indications for reduction, conception mode, nulliparous As FR in twins is increasingly practiced, counseling couples with the timing of reduction and risks should be based on reliable data. Therefore, we present our data on pregnancy outcomes in a longitudinal cohort of DC twin pregnancies reduced to singletons and graphically depict the relationship between the timing of reduction and pregnancy outcome. The curve fitting revealed a significant turning point (17 +4  weeks) of GA at reduction for premature birth risk in DC twin pregnancies. Specifically, the negative effects of GA at reduction on GA at birth and birthweight were only observed for reduction performed after 17 +4  weeks. The rates of preterm birth < 32 weeks or < 34 weeks for reduction performed before 17 +4  weeks were approximately a quarter of that for reduction performed after 17 +4  weeks. Additionally, the rates of preterm birth < 37 weeks and LBW for reduction performed after 17 +4  weeks were nearly three times higher than those for reduction performed before 17 +4  weeks. On average, the birthweight in reductions performed after 17 +4  weeks was 314 g lower than that of reductions performed before 17 +4  weeks. However, no correlation between GA at reduction and pregnancy loss risk was found. To date, questions still remain about the relationship between GA at reduction and the risk of premature birth. Previous studies yielded inconsistent findings, largely due to varying grouping methods and relatively small samples (Table  5 ). Yaron et al. [ 27 ] found earlier delivery and lower birthweight in twins reduced from 14 weeks in a cohort of 82 DC twins, which is consist with a recent study of 172 DC twins by Steffen [ 30 ]. Zemet et al. [ 24 ] found reduction performed from 15 weeks increased risk of preterm birth and LBW in a cohort of 248 DC twins. Alvarado et al. [ 23 ] revealed a decrease in birthweight for reductions performed from 18 weeks in a cohort of 28 DC twins. Lynch et al. [ 26 ] suggested that reductions performed from 20 weeks was associated with a worse perinatal outcome in a cohort of 69 DC twins. In contrast, Eddleman et al. [ 31 ] found no difference in premature delivery risk of reduction in multiple pregnancies before or from 20 weeks. In this study, we identified the turning point at the GA of reduction (17 +4  weeks) that influences fetal prematurity for the first time. A dramatic decrease in GA at birth and birthweight was shown when reduction was performed after 17 +4  weeks. On average, for each week of delay in performing the reduction after 17 +4  weeks, for each week of delay, the GA at birth is reduced by 0.51 weeks and birthweight is decreased by 100.35 g. This delay is associated with significantly higher risks of preterm birth (< 32, < 34, or < 37 weeks) and LBW. However, our findings indicate that the GA at birth and birthweight were not significantly related to the timing of reduction when it was performed before 17 +4  weeks. These findings highlight that performing the reduction before 17 +4  weeks could decrease prematurity-related adverse outcomes. Table 5 Association between timing of fetal reduction and pregnancy outcome in previously published studies GA at FR Multiple or twins Indications for reduction N Pregnancy loss (n%) P value Preterm delivery (n%) P value Evans et al.1994 [ 20 ]  ≤ 16w multiple pregnancies selective termination 37 2(5.4%)  < 0.01   16w 146 21(14.4%) 54(43.2%) Lynch et al.1996 [ 25 ]  ≤ 20w twin pregnancies selective termination 28 1(3.6%) 0.99   20w 41 1(2.4%) 20(50%) Yaron et al.1998 [ 26 ]  ≤ 14w DC twins selective termination and elective reduction 31 3 (9.7%) NS   14w 51 4 (7.8%) 13 (27.6%) Evans et al.1999 [ 18 ] 9-12w multiple pregnancies selective termination 402 5.40% Not available 13-18w 8.70% 19-24w 6.80%  > 24w 9.10% Eddleman et al.2002 [ 30 ]  ≤ 20w multiple pregnancies selective termination 119 7(5.9%) 0.09   20w 79 1(1.3%) 2 (2.6) Alvarado et al.2012 [ 22 ]  < 18 w DC twins selective termination 12 0 0.1  ≤ 34w 2(16.7%) 0.1  ≥ 18w 16 1(6.25%) 2(13.3%) Bigelow et al.2014 [ 21 ]  < 18w DC twins selective termination 47 0 0.05 Not available  ≥ 18w 33 4(12.1%) Kim et al.2019 [ 19 ]  < 15w DC twins (i) the presence of genetic or congenital anomaly (n = 23); 35 1(2.9%) NS  < 37w 4(11.8%) NS  ≥ 15w (ii) an obstetrical indication specified as a past maternal history of preterm delivery that caused fetal death or cerebral palsy of the child (n = 21) 9 0 3(33.3%) Kim et al.2019 [ 24 ] 6-8w multiple pregnancies Triplets, higher-order pregnancy, severe fetal anatomical abnormality determined by ultrasonography, chromosomal abnormality, poor obstetric history and miscellaneous indications 181 16 (8.8%) 0.045  < 37w 89(49.2%) NS 11-20w 115 3 (2.6%) 49(42.6%) Zemet et al.2021 [ 23 ] 11–14w DC twins Structural or genetic abnormalities 172 1 (0.6%) 0.52  < 37w 24 (14.0%) 0.01 15–23w Mainly at the patient’s request or because of an unfavorable obstetric or maternal medical history, with a minority undergoing selective reduction 76 1 (1.3%) 21 (28.0%) Steffen et al.2023 [ 29 ]  < 14w DC twins Not available 73 1(1.4%) Not available  < 37w 5(6.9%) Not available  ≥ 14w 99 5(5.1%) 18(19.4%) Abbreviations : GA  Gestational age, FR  Fetal reduction, DC  Dichorionic Association between timing of fetal reduction and pregnancy outcome in previously published studies Abbreviations : GA  Gestational age, FR  Fetal reduction, DC  Dichorionic It remains a controversial issue as to whether the timing of reduction affects the risk of pregnancy loss (Table  3 ). Evans et al. [ 21 ] found a higher pregnancy loss risk in reduction performed from 16 weeks in multiple pregnancies, while other studies suggested GA at reduction did not correlate with the risk of pregnancy loss [ 20 , 23 , 24 , 26 , 27 ]. Consisted with most previous studies, our study found that pregnancy loss before 24 weeks was not associated with GA at reduction and the risk was similar to that reported by the previous studies [ 13 , 18 , 32 , 33 ]. Our results support the majority of previous published reports that performing reduction at a later GA in DC twins increases the risk of preterm birth and low birthweight [ 23 , 24 , 27 , 28 ]. Importantly, we found an alert timeline for reduction for negative effects on pregnancy outcomes. From the turning point 17 +4  weeks, the later the reduction was performed in DC twins, the earlier the delivery and lower the birthweight. Our study revealed that in the Early-FR group, only 24.3% of patients experienced preterm birth, whereas in the Late-FR group, 75% of patients had preterm deliveries. This significant discrepancy may also be the primary reason for the observed differences in birth weight between the two groups. In the past, most FRs of twins were performed for termination of abnormal fetus. Recently, elective FR of twins is increasingly practiced in couples consider to optimize the pregnancy outcome because of poor obstetric history, adverse uterine factors, or other miscellaneous factors [ 18 – 20 ]. In this study cohort, the most common indications for FRs were adverse uterine factors and miscellaneous factors, while anatomical abnormality or chromosomal anomaly account for only a small proportion. Evaluation of fetuses before reduction leads to improved outcomes. Our findings suggest that in cases of FRs due to anatomical or chromosomal abnormalities, performing FRs as early as possible (preferably within 17 +4  weeks) may guarantee a lower risk of preterm delivery once this condition is confirmed. However, while some fetal anomalies are not clarified until a second fetal scan is performed [ 34 , 35 ] or a CVS is performed when a genetic disorder is suspected [ 36 ], later FR will still be necessary. Women requesting for elective FR are quite different from women who choose to undergo selective termination for known fetal malformations. Deciding when to performed an elective FR is always a challenging decision. A common concern is the increased risk of premature birth associated with FR at later gestational weeks. Although FR at early GA is more acceptable from religious and ethical perspectives, the loss of one fetus in a twin pregnancy or spontaneous miscarriage always occurs in the first trimester, and fetal anomalies are rarely detectable by ultrasonography at an early GA [ 36 ], therefore it might be too early to perform the reduction at this time. Our study findings did not suggest any obvious benefits of performing FRs at a very early stage. Therefore, we recommend performing FR after a thorough ultrasonographic assessment to minimize the risk of reducing a normal fetus and retaining an abnormal one, especially in high-risk populations. Earlier diagnosis of any aneuploidy or fetal malformations is particularly important in twins. The ultrasound, and CVS are able to help decide which to keep or reduce. Although NIPT cannot distinguish which fetus is which, if an abnormality is detected, diagnostic techniques like CVS will still be required. Technological advances have made it possible to detected many major fetal anomalies at the first-trimester screening and early second trimester of pregnancy [ 37 , 38 ]. Combined with the findings from our study, FR before 17 +4  weeks might result in a relatively favorable outcome. Monitoring and management of premature related risks should be taken place for twins reduced after 17 +4  weeks. Our results need to be further validated in multicenter cohort studies. Although we acknowledge that FR will never be widely accepted, we contend that it should be viewed as a reasonable procedure in some situations that, ideally, becomes more uncommon from the standpoints of autonomy and public health. These findings could provide a basis for clinical practice and add information to the pros and cons of when to FR in a consultation [ 39 ]. Two common FR methods were performed in our study, as transvaginal FR was performed at 6 +5 –11 +6  weeks, and transabdominal FR was performed at 13 +1 –26 +0  weeks. Because the choice of FR procedural methods and the gestational week at reduction are closely related, these two strongly covariant factors cannot be included in the same model simultaneously. Therefore, using curve fitting to determine differences before and after 12–13 weeks (the demarcation between transvaginal and transabdominal FR) is a good way to determine the impact of the mode of fetal reduction. In this study, no turning point near 12–13 weeks was identified in the curves, suggesting that the different methods of FR are not a major factor affecting pregnancy outcomes. Given these findings, the conclusions regarding the differences of two methods drawn in some previous studies may warrant further investigation. Importantly, 81.4% twin pregnancies in our study were concepted from IVF/ICSI. As our previous study indicated that singletons reduced from twins still had higher rates of preterm birth, LBW, and SGA than primary singletons [ 16 ], the best way to prevent adverse pregnancy outcome related with premature is limiting the number of transferred embryos and the advocating of SET [ 40 ]. To our knowledge, this study represents the largest cohort of DC twin pregnancies undergoing FRs, and we adjusted for a number of important maternal confounders to analysis. Previous studies focusing on the timing of reduction have typically used different gestational weeks as criteria for grouping, rather than identifying the actual inflection point, which may not be a completely true reflection of how pregnancy outcomes change with GA at reduction, as we often encounter non-linear data, such as the appearance of a U-shaped curve. In this situation, blindly using linear regression or direct grouping may hide the relationship between the two variables, making us miss the"real culprit". The major strength of this study is that it is the first time to use smooth curve fitting and threshold analysis to find the change pattern and the most exact timing boundary of reduction. Our findings may provide more detail guidance to patients and physicians considering reduction of DC twin pregnancies. This study has some limitations. First, the fetal reduction cohort in this study spans a relatively long timeframe. However, the procedural protocol for fetal reduction remained consistent, with all procedures performed by the same team of experienced operators during the period. We believe that the technical homogeneity of the procedures minimizes potential variability. Second, although our cohort is large compared with published literature, the sample size in late-reduction group was relatively small. The retrospective nature of the study limits causal inference and introduces potential biases. Future prospective studies would strengthen the results. Additionally, since our study is not a randomized control trial (RCT), the analysis is subject to unescapable confounding and missing data bias. However, the ethical challenges make it difficult to conduct RCTs in this area.

Four hundred seventy-five women with DC twin pregnancies who underwent FR between July 2005 and July 2021 at the reproductive center, Women’s Hospital, Zhejiang university, School of medicine were enrolled in this cohort study. The presence of two separate placentas or the presence of a “lambda sign” of the intertwin membrane confirms as dichorionicity. Exclusion criteria were pregnancy loss before FR, undergoing FR for the second time in the present pregnancy or incomplete data for both the mother and neonate. A total of 463 twin pregnancies underwent FR were included in the study after excluding pregnancy loss before FR ( n  = 2), undergoing FR for the second time in the present pregnancy ( n  = 3), and incomplete follow-up data ( n  = 7) (Fig.  1 ). 350 women underwent transvaginal FR at 6 +5 –11 +6  weeks, and 113 women underwent transabdominal FR at 13 +1 –26 +0  weeks after first trimester scan. The indications for FR were sonographic anatomical or chromosomal abnormalities, a poor obstetric history, adverse uterine factors and miscellaneous indications. Adverse uterine factors included large uterine scar, severe adenomyosis and uterine malformation. Miscellaneous indications refer to factors that contributed to inability to bear multiple fetuses, such as psychological, socioeconomic reasons, short stature, unfavorable maternal medical history et al., or purely to reduce the number of fetuses aimed at improving outcomes. A"poor obstetric history"is defined as a history of one or more of the following adverse pregnancy outcomes in previous pregnancies, including preterm delivery, preterm premature rupture of membranes (PPROM), miscarriage in second trimester. The Institutional Review Board (IRB) of Women's Hospital, School of Medicine, Zhejiang University, approved the study (approval number: 20181077). Written informed consent was obtained from all patients. Fig. 1 Flowchart of cohort and the gestational age distribution of fetal reduction for different indications Flowchart of cohort and the gestational age distribution of fetal reduction for different indications All patients underwent counselling regarding the risks and benefits of FR. FR procedures were performed in transvaginal or transabdominal methods by highly skilled physicians according to the same protocol [ 15 ]. The abnormal fetal or the smallest fetal, or fetal located in a position with the easiest access route was selected for reduction. Transvaginal FR was performed as previously described [ 16 ]. Briefly, a needle of appropriate size was inserted into the fetal heart to aspirate the fluid and fetus from the sac (17G size needle) or inject potassium chloride solution (0.6–2.0 ml) (21G size needle). The procedure for transabdominal FR was performed after first trimester screening. Under the guidance of an abdominal ultrasound transducer, a 22G needle was inserted into the uterine cavity and then into the fetal heart or thoracic cavity, an intracardiac or intrathoracic injection of potassium chloride (1.5–7.5 ml) was administered. All patients underwent an ultrasound examination shortly after the procedure and after one week to confirm the viability of the remaining fetus. Baseline characteristics and clinical data were obtained from electronic medical records. Pregnancy outcomes assessed in this study included GA at birth, mode of delivery, birthweight, and the rates of pregnancy loss before 24 weeks, stillbirth from 24 weeks, live birth, preterm birth, low birthweight (LBW), small for gestational age (SGA), and pregnancy complications, including hypertensive disease of pregnancy, gestational diabetes mellitus, and preterm premature rupture of membranes (PPROM) were determined. LBW was defined as birth weight below 2500 g. SGA was defined as birthweight below the 10th percentile for the GA at birth using the Chinese birthweight reference [ 29 ]. We adjusted for maternal age, height and pre-conception BMI, indications for reduction, conception mode, nulliparous to control for the effects of the potential confounders both in smooth curve fitting and Logistic regression analysis. Since the choice of FR procedural methods (transvaginal FR vs. transabdominal FR) and the GA at reduction are closely related, these two strongly covariant factors cannot be placed into a model at the same time. Therefore, the FR procedural methods were not included in the adjusted model. Instead, we used curve fitting to determine if there was a difference before and after 12–13 weeks (the demarcation between transvaginal and transabdominal FR) is a good way to determine the impact of the mode of fetal reduction. Continuous variables are presented as Mean and ± SD, and categorical variables are presented as N (%). Smooth curve fitting was conducted to see whether there were nonlinear relationships between GA at reduction and pregnancy loss risk, GA at birth, and birthweight. A segmented regression model and Log-likelihood ratio test (LRT test) were used to analyze the threshold effect between GA at reduction and those variables and to identify the inflection points (i.e., turning point), which is using a separate line segment to fit each interval. Smooth curve fitting and segmented regression model were performed with EmpowerStats ( http://www.empowerstats.com ) and R-project. The statistical significance of the differences between two groups were analyzed using the χ2 test, Fisher’s exact test, the Mann–Whitney U -test, and two-tailed unpaired Student’s t- test, as appropriate. Logistic regression analysis was used for the multivariate analysis while controlling for the effects of potential confounders. Statistical analysis was performed using SPSS Statistics software, version 26 (IBM Corp., Armonk, NY, USA). A P -value < 0.05 was considered statistically significant.

This study indicates that pregnancy loss risk was not associated with GA at reduction. Fetal reduction before 17 +4  weeks is overall a better option for DC twin pregnancies with a lower risk of prematurity. GA at birth and birthweight were negatively correlated with GA at reduction when the reduction is performed after 17 +4  weeks. Thus, for twins reduced after 17 +4  weeks, monitoring and management of premature related risks should be taken place.

The incidence of multiple pregnancies has risen over the last few decades, mainly attributed to the use of assisted reproduction techniques (ART), and ovulation induction drugs [ 1 , 2 ]. The promulgation of some national guidelines that recommend reducing the number of embryos transferred and controlling the use of ovulation induction drugs has led to a decline in the incidence of higher-order multiple pregnancies [ 3 ]. However, a recent report indicates that the global twinning rate has increased by a third, from 9.1 to 12.0 twin deliveries per 1000 deliveries, accounting for approximately 1.6 million twin pairs annually [ 4 ]. Meanwhile, the twin birth rate in China increased from 2.84% in 2012 to 3.22% in 2020 [ 5 ]. Twin pregnancies are classified as a high-risk pregnancies, as the risk of preterm birth < 34 weeks gestation in twin pregnancies exceeds 50% [ 6 ], along with an increased risk of maternal complications and offspring morbidity [ 7 – 9 ]. Fetal reduction (FR) is a secondary method for improving maternal and perinatal outcome that reduces the total number of fetuses by one or more in a multiple pregnancy [ 10 , 11 ]. In recent years, the vast majority of studies, including our previous study, have concluded that reducing twins to singletons is effective in improving pregnancy outcomes in terms of prematurity and birthweight without increasing pregnancy loss risk [ 12 – 16 ]. In the past, selective termination which that refers to reduction of fetus with genetic or structural abnormality was the main reason for twins reduction [ 17 ]. Recently, there is an increasing number of couples with twin pregnancies have opted for elective reduction to decrease risks associated with maternal conditions such as a poor obstetric history, adverse uterine factors, known complications of twin pregnancies or for socioeconomic reasons [ 18 – 20 ]. There is growing concern about the timing of reduction, that may influence the risk of pregnancy loss or premature birth with limited evidence. Previous studies have inconsistent conclusions regarding the optimal timing for fetal reduction in dichorionic (DC) twin pregnancies to minimize the risk of adverse pregnancy outcomes[ 17 , 20 – 28 ]. Meanwhile, these studies are limited by relatively small sample sizes and different subgroup comparisons. Therefore, this study aimed to investigate the association between gestational age (GA) at reduction in DC twin pregnancies and risk of adverse pregnancy outcomes, in a relatively large cohort. Smooth curve fitting and a segmented regression model were performed to identify the most exact timing boundary of reduction.

Supplementary Material 1 Supplementary Material 1