Predictive value of a short cervix in the early third trimester for preterm birth in twin pregnancies Authors: Oyoung Kim, Hong Yeon Lee, Hyun Sun Ko, In Yang Park, Subeen Hong Published: 2026 ABSTRACT AbstractBackgroundPreterm birth (PTB) is a major cause of neonatal morbidity and mortality, and cervical length measured in the mid-trimester is a well-established predictor of PTB. However, the predictive value of cervical length in the early third trimester remains unclear, particularly in twin pregnancies. This study aimed to investigate the association between a short cervical length (≤ 25 mm) measured during the early third trimester (28 + 0 to 33 + 6 weeks) and the risk of PTB before 37 weeks, and to evaluate its predictive utility.MethodsThis retrospective cohort study included women with twin pregnancies who delivered at a single tertiary hospital between 2009 and 2019. Women with a history of cervical cerclage, iatrogenic preterm birth, or missing cervical length measurements in the second or early third trimester were excluded. Participants were categorized into short cervix and control groups based on cervical length in the early third trimester.ResultsAmong 291 twin pregnancies, 102 (35.1%) resulted in PTB. The incidence of PTB was significantly greater in the short cervix group (73.0%) compared to the control group (29.0%, P < 0.001). A short cervix in the early third trimester was associated with a 3.7-fold increased risk of PTB. Among women with a normal cervical length in the second trimester, those with cervical shortening in the early third trimester had a 2.3-fold increased risk. Incorporating third-trimester cervical length improved PTB risk prediction, increasing the area under the receiver operating characteristic curve from 0.629 to 0.711 (P = 0.001).ConclusionsA short cervical length in the early third trimester is associated with an increased risk of PTB in twin pregnancies. Assessing cervical length during this period may improve risk stratification and inform clinical assessment, particularly among asymptomatic or relatively lower-risk populations.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12884-026-09302-y.Keywords: Cervical length measurement; Pregnancy trimester, third; Pregnancy, twin; Pregnancy trimester, second; Risk assessment Background Preterm birth (PTB) is a leading cause of neonatal morbidity and mortality, presenting significant challenges in perinatal care [1, 2]. The risk of PTB is particularly high in twin pregnancies, with nearly 50% of births occurring before 37 weeks of gestation [3, 4]. As maternal age increases and the use of assisted reproductive technologies continues to expand, the incidence of multiple gestations is also increasing. This trend highlights the need for reliable predictors of PTB in twin pregnancies [5–8]. Identifying reliable predictors of PTB is important for timely and effective interventions to improve neonatal outcomes. Transvaginal ultrasound measurement of cervical length is a safe and reliable tool for assessing PTB risk [9–11]. However, while mid-trimester cervical length screening is well established to have predictive value, there is limited research on the relationship between cervical length measured in the early third trimester and risk of spontaneous PTB, particularly in twin pregnancies [12–14]. Moreover, a short cervical length in the early third trimester is often considered a normal physiological change, but evidence to support this assumption is lacking [15]. This study investigated the association between a short cervical length (≤ 25 mm) measured in the early third trimester (28 + 0 to 33 + 6 weeks) and risk of PTB in twin pregnancies. Additionally, it aimed to evaluate the predictive value of early third-trimester cervical length measurements for PTB and their potential clinical utility. Methods Study design and population This retrospective cohort study collected delivery information and prenatal ultrasound records for all women who delivered twins at a single tertiary university hospital between 2009 and 2019. Women were excluded if they had received cervical cerclage, experienced an iatrogenic PTB, or lacked documented cervical length measurements in either the second or third trimester. Spontaneous PTB was defined as preterm delivery due to spontaneous preterm labor or preterm premature rupture of membranes (PPROM). In cases with mixed or unclear indications, classification was based on the primary clinical indication; cases were classified as spontaneous PTB when preterm labor or PPROM was the primary indication rather than an iatrogenic cause. The study population was categorized into two groups, a short cervix group (cervical length ≤ 25 mm) and a control group (cervical length > 25 mm), based on early third-trimester cervical length. The present study investigated the association between short cervical length in the early third trimester and PTB and then evaluated the predictive value of third-trimester cervical length measurements for spontaneous PTB. In a subanalysis, women with a second-trimester cervical length > 25 mm were included. This approach evaluated the association of a short cervix in the early third trimester with a previously normal cervical length with increased risk of spontaneous PTB. This study was approved by the Institutional Review Board (IRB) of Seoul St. Mary’s Hospital (IRB no. KC23RISI0613). Assessment of cervical length At the study institution, routine cervical length measurements during prenatal ultrasound examinations are standard antenatal care for women attending outpatient clinics [16]. Pregnant women typically attend prenatal visits monthly until 28 weeks of gestation, every two weeks from 28 to 36 weeks of gestation, and weekly thereafter. For study purposes, certified sonographers accurately measured cervical length in accordance with established protocols. Briefly, cervical length measurements were performed via transvaginal ultrasound in a midsagittal plane, ensuring an empty bladder and avoiding excessive pressure. Measurements were performed between the internal and external os, and the shortest length of three measurements was recorded. Cervical length measurements obtained during emergency department visits or hospitalizations due to urgent symptoms, such as preterm labor or preterm pre-labor membrane rupture, were not included in the study. For analysis, if multiple cervical length measurements were available within a given gestational period (16 + 0 to 27 + 6 weeks or 28 + 0 to 33 + 6 weeks), the shortest measurement was selected to represent that period. Maternal characteristics and pregnancy outcomes Maternal characteristics, including age, gravidity, parity, chorionicity, and medical and obstetric histories, as well as delivery details and pregnancy and neonatal outcomes, were obtained through a retrospective review of medical records. At our institution, vaginal progesterone was prescribed for women with cervical shortening (cervical length ≤ 25 mm), based on institutional practice patterns and individual physician judgment. The primary outcome was spontaneous PTB at < 37 weeks of gestation. Patients were subdivided into early PTB cases, defined as those occurring before 34 weeks 0 days, and late PTB cases, categorized as occurring from 34 weeks 0 days to 36 weeks 6 days. Statistical analysis Continuous data are described as median with interquartile range (IQR), while categorical data are presented as frequencies and percentages. Group differences were evaluated using the Chi-square test or Fisher’s exact test for categorical variables and the Wilcoxon rank-sum test for continuous variables. Kaplan–Meier survival curves were generated to estimate the cumulative incidence of PTB based on the presence or absence of a short cervix in the third trimester. For repeated-measures analysis, linear mixed-effects regression models with a random intercept and an unstructured covariance matrix were employed to assess changes in cervical length over time and to detect differences between the PTB group and the term birth group, with the results illustrated by least squares means and corresponding error bar plots. We performed univariable and multivariable logistic regression analyses to explore the associations among clinical factors, short cervix, and risk of PTB. To evaluate the additional predictive value of cervical length factors to clinical factors, two multivariable models were developed. The selection of variables was based on a univariable analysis criterion of P < 0.25. The base model included maternal age, history of PTB, chorionicity, use of vaginal progesterone, and cervical length measured before 28 weeks, while the cervical length–adjusted model additionally incorporated the short cervix group variable, defined by a short cervical length in the early third trimester. To address potential concerns regarding multicollinearity, cervical length measurements from the second and third trimesters were not included simultaneously in the regression models. Cervical length was treated as a categorical variable based on the presence of a short cervix (≤ 25 mm) during the third-trimester window. Sensitivity analyses were performed by categorizing patients according to cervical length status (≤ 25 mm vs. >25 mm) in each trimester to evaluate combined effects across trimesters. An alternative cutoff (≤ 20 mm) was additionally applied to the regression models to assess the robustness of the findings. The discriminatory performance of the models was evaluated using the area under the receiver operating characteristic curve (AUC), with differences between the models compared according to DeLong’s method. Further, the additional predictive performance of cervical length was assessed by calculating changes in the continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI) metrics. All P values were calculated using two-sided tests, with statistical significance defined as P < 0.05. Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA). Results Characteristics study population and delivery outcomes Among the 697 women who delivered twins during the study period, 45 underwent cervical cerclage, 133 were delivered for iatrogenic indications, and 203 lacked cervical length measurements in either the second or third trimester. Another 25 pregnancies had cervical length measurement only in the second trimester and also were excluded, resulting in a final study population of 291 twin pregnancies. These 291 pregnancies were ultimately included in the study, with 141 (48.4%) placed in the short cervix group and 150 (51.5%) placed in the control group (Fig. 1). To assess potential selection bias, baseline characteristics of the included study population were compared with those of the excluded population and are presented in Supplementary Table 1. Study flow diagram No significant differences were observed between the short cervix group and the control group in terms of mean maternal age, nulliparity, pre-pregnancy body mass index (BMI), chorionicity, in vitro fertilization (IVF), prior loop electrosurgical excision procedure (LEEP), prevalence of placenta abnormalities, gestational diabetes mellitus, or history of PTB (Table 1). The short cervix group had a lower incidence of pregnancy-associated hypertension (1.4% vs. 6.7%, P = 0.024) and a lower prevalence of smoking history (P = 0.037) compared to the control group. The number of cervical length assessments and the gestational age at measurement after 28 weeks did not differ significantly between the groups. After 28 weeks, the short cervix group had a mean cervical length of 1.8 ± 0.5 cm, whereas the control group had a mean cervical length of 3.8 ± 0.5 cm. The short cervix group experienced an earlier mean gestational age at birth and a lower neonatal birth weight compared to the control group. The overall incidence of spontaneous PTB was significantly higher in the short cervix group (73.0%) compared to the control group (29.0%, P < 0.001). Additionally, both early PTB and late PTB were more prevalent in the short cervix group, with early PTB occurring in 14.0% of cases compared to 4.0% in the control group (P = 0.001) and late PTB occurring in 41.8% of cases compared to 16.7% in the control group (P < 0.01). Figure 2 shows the Kaplan–Meier curve illustrating the cumulative incidence of PTB by gestational age. The short cervix group exhibited a significantly higher PTB rate compared to the control group, with a marked increase after 32 weeks (log-rank test, P < 0.001). Table 1Baseline characteristics, cervical length, and delivery outcomes of the study populationTotal(n = 291)Short cervix group(n = 141)Control group(n = 150) P valueMaternal baseline characteristicsMaternal age (year)34.0 (31.0–36.0)33.0 (31.0–36.0)34.0 (31.0–36.0)0.620Nulliparity243 (83.5)123 (87.2)120 (80.0)0.097History of preterm birth3 (1.0)0 (0.0)3 (2.0)0.248Pre-pregnancy BMI (kg/m2)17.7 (15.2–20.5)17.5 (15.0-20.1)17.8 (15.3–21.1)0.349IVF141 (48.5)71 (50.4)70 (46.7)0.529Chorionicity0.711 Dichorionic264 (90.7)127 (90.1)137 (91.3) Monochorionic27 (9.3)14 (9.9)13 (8.7)Prior LEEP2 (0.7)1 (0.7)1 (0.7)> 0.999Smoking9 (3.1)1 (0.7)8 (5.3)0.037Drinking10 (3.4)2 (1.4)8 (5.3)0.105Placenta abnormalities> 0.999 None288 (99.0)140 (99.3)148 (98.7) Placenta previa2 (0.7)1 (0.7)1 (0.7) Placenta abruptio1 (0.3)0 (0)1 (0.7)Pregnancy associated hypertension12 (4.1)2 (1.4)10 (6.7)0.024Gestational diabetes mellitus22 (7.6)11 (7.8)11 (7.3)0.880Factors associated with cervical lengthVaginal progesterone use22 (7.6)18 (12.8)4 (2.7)0.001 GA at initiation of vaginal progesterone (weeks) (n = 22)28.2 ± 3.827.6 ± 3.931.0 ± 0.80.003 Duration of vaginal progesterone (days) (n = 22)40.7 ± 25.744.7 ± 26.522.8 ± 9.60.124Number of checking CL4.0 (2.0–5.0)4.0 (2.0–5.0)4.0 (2.0–5.0)0.459GA at measuring CL before 28 weeks (n = 200)25.0 (22.0–26.0)25.0 (23.5–26.5)24.0 (21.0–26.0)0.035CL before 28 weeks (cm) (n = 200)3.4 ± 0.73.1 ± 0.73.8 ± 0.5< 0.001GA at measuring CL after 28 weeks (weeks)32.0 (30.0–33.0)32.0 (30.0–33.0)32.0 (30.0–32.0)0.214CL after 28 weeks (cm)2.6 ± 0.91.8 ± 0.53.3 ± 0.6< 0.001Delivery outcomesGA at birth (weeks)37.0 (36.0–37.0)36.0 (35.0–37.0)37.0 (37.0–38.0)< 0.001Overall preterm birth102 (35.1)73 (51.8)29 (19.3)< 0.001Early preterm birth18 (6.2)14 (9.9)4 (2.7)0.010Late preterm birth84 (28.9)59 (41.8)25 (16.7)< 0.001Cesarean section delivery255 (87.6)120 (85.1)135 (90.0)0.205Birth weight (kg)2.5 (2.3–2.8)2.5 (2.2–2.7)2.6 (2.4–2.8)0.006Data are presented as median (interquartile range) or as number (percentage)Abbreviation: BMI body mass index, IVF, in vitro fertilization, LEEP loop electrosurgical excision procedure, CL cervical length, GA gestational ageP values were calculated using the Chi-square, Fisher’s exact test, and Wilcoxon rank-sum test Baseline characteristics, cervical length, and delivery outcomes of the study population Data are presented as median (interquartile range) or as number (percentage) Abbreviation: BMI body mass index, IVF, in vitro fertilization, LEEP loop electrosurgical excision procedure, CL cervical length, GA gestational age P values were calculated using the Chi-square, Fisher’s exact test, and Wilcoxon rank-sum test Kaplan–Meier plot of cumulative preterm birth incidence birth. The short cervix group (red line) showed a higher preterm birth rate after 32 weeks compared with the control group (blue line) Association between short cervical length in the third trimester and risk of preterm birth Table 2 presents the findings of univariable and multivariable logistic regression analyses for PTB, expressed as odds ratios. Factors significantly associated with PTB were maternal age, chorionicity, short cervix group, use of vaginal progesterone, and cervical length before 28 weeks. The base model incorporated these variables, along with a history of PTB, and which is a well-established risk factor. The cervical length–adjusted model additionally included the short cervix variable to assess its independent contribution to PTB risk. Notably, a short cervical length after 28 weeks was associated with a 3.7-fold increased risk of PTB. Per the least squares means shown in Fig. 3, both PTB and advanced gestational age were significantly associated with a shortened cervical length. However, the rate of cervical length decrease over time did not differ notably between the two groups. A subgroup analysis was performed including 179 women with a short cervix after 28 weeks, which remained a significant predictor among those with a cervical length > 25 mm before 28 weeks (Fig. 1). A comparison of baseline characteristics, cervical lengths, and delivery outcomes between the short cervix group and the control group and the results of univariable and multivariable logistic regression analyses for PTB are presented in Supplementary Tables 2 and 3, respectively. Of particular significance, even among women with a cervical length > 25 mm in the second trimester, those who subsequently developed a short cervix after 28 weeks experienced a 2.3-fold increase in the risk of PTB. In sensitivity analyses assessing the combined effect of cervical length across trimesters, the rate of preterm birth was highest among women with a short cervix in both trimesters (70.0%) and lowest among those with normal cervical length in both trimesters (24.3%). Women with a normal cervical length in the second trimester but a short cervix in the third trimester also showed an increased rate of preterm birth (47.4%). Table 2Univariable and multivariable logistic regression analyses for preterm birthTerm birthPreterm birthBase ModelCL-adjustedModeln = 189n = 102crude OR(95% CI) P valueadjusted OR(95% CI) P valueadjusted OR(95% CI) P valueMaternal baseline characteristicsMaternal age (year)34.0 (32.0–36.0)33.5 (31.0–36.0)0.95 (0.88–1.02)0.1350.94 (0.87–1.01)0.1050.94 (0.87–1.02)0.141Nulliparity158 (83.6)85 (83.3)0.97 (0.51–1.85)0.929History of preterm birth2 (1.1)1 (1.0)1.11 (0.11–11.6)0.9311.98 (0.17–23.5)0.5883.20 (0.28–36.7)0.349Pre-pregnancy BMI (kg/m2)17.7 (15.3–20.9)17.7 (14.9–20.2)0.99 (0.95–1.02)0.483IVF90 (47.6)51 (50.0)1.10 (0.68–1.78)0.700Monochorionic13 (6.9)14 (13.7)2.14 (0.97–4.75)0.0611.80 (0.78–4.13)0.1672.03 (0.85–4.84)0.110Prior LEEP1 (0.5)1 (1.0)1.86 (0.11–29.98)0.663Placenta previa1 (0.5)1 (1.0)1.85 (0.11–29.85)0.666Placenta abruptio1 (0.5)0 (0.0)0.59 (0.01–57.71)0.820Smoking5 (2.6)4 (3.9)1.53 (0.40–5.83)0.531Drinking6 (3.2)4 (3.9)1.29 (0.36–4.66)0.697Pregnancy associated hypertension9 (4.8)3 (2.9)0.67 (0.18–2.45)0.544Gestational diabetes mellitus13 (6.9)9 (8.8)1.33 (0.55–3.22)0.529Factors associated with cervical lengthShort cervix group68 (36.0)73 (71.6)4.42 (2.62–7.45)< 0.0013.70 (2.09–6.56)< 0.001Control group121 (64.0)29 (28.4)referenceVaginal progesterone9 (4.8)13 (12.7)2.87 (1.18–6.95)0.0202.46 (0.98–6.17)0.0541.89 (0.74–4.84)0.183Number of checking CL4.0 (2.0–5.0)4.0 (2.0–5.0)1.01 (0.89–1.13)0.919GA at measuring CL before 28 weeks (weeks) (n = 200)23.7 ± 3.024.0 ± 3.01.04 (0.94–1.14)0.449CL before 28 weeks (cm) (n = 200)3.6 ± 0.63.2 ± 0.70.46 (0.29–0.73)0.0010.48 (0.30–0.78)0.0030.75 (0.45–1.25)0.267Data are presented as median (interquartile range) or as number (percentage). P values were calculated using univariable and multivariable logistic regression analyses of the baseline factors for preterm birth. The adjusted covariates were maternal age, history of preterm birth, vaginal progesterone use, and cervical length before 28 weeks, each of which had a P-value < 0.25 in the univariable analysis. The base model included covariates of maternal age, history of preterm birth, chorionicity, vaginal progesterone use, and cervical length before 28 weeks. The cervical length-adjusted model included all covariates from the base model with the addition of the short cervix variableAbbreviation: BMI body mass index, IVF in vitro fertilization, LEEP loop electrosurgical excision procedure, CL cervical length, GA gestational age Univariable and multivariable logistic regression analyses for preterm birth Data are presented as median (interquartile range) or as number (percentage). P values were calculated using univariable and multivariable logistic regression analyses of the baseline factors for preterm birth. The adjusted covariates were maternal age, history of preterm birth, vaginal progesterone use, and cervical length before 28 weeks, each of which had a P-value < 0.25 in the univariable analysis. The base model included covariates of maternal age, history of preterm birth, chorionicity, vaginal progesterone use, and cervical length before 28 weeks. The cervical length-adjusted model included all covariates from the base model with the addition of the short cervix variable Abbreviation: BMI body mass index, IVF in vitro fertilization, LEEP loop electrosurgical excision procedure, CL cervical length, GA gestational age Least squares mean plot with 95% confidence intervals. Preterm birth status and advancing gestational age were associated with shorter cervical length. However, the rate of cervical length decrease over time did not differ significantly between the two groups. Results are shown for mixed model (A) without covariates and (B) with covariates Predictive performance of cervical length in the third trimester To evaluate whether the consideration of short cervix status in the early third trimester improves the prediction of PTB, model performance was assessed using the AUC, IDI, and NRI metrics (Table 3). In the primary analysis of the entire study cohort, incorporating short cervix status into the base model significantly enhanced predictive performance. The AUC increased from 0.637 (95% confidence interval [CI], 0.569–0.706) to 0.716 (95% CI, 0.653–0.780) (P = 0.001). Additionally, the IDI was 0.071 (95% CI, 0.043–0.100; P < 0.001) and the NRI was 0.594 (95% CI, 0.365–0.824; P < 0.001), respectively, indicating significant improvements in both model discrimination and classification. Table 3Assessment of the incremental value of third-trimester short cervix for preterm birth predictionAnalysisModelAUC(95% CI)AUC difference (95% CI) P valueIDI(95% CI) P valueNRI(95% CI) P valuePrimaryanalysisBase Model0.637 (0.569–0.706)referenceCL-adjusted Model0.716 (0.653–0.780)0.079 (0.031–0.127)0.0010.071 (0.043-0.100)< 0.0010.594 (0.365–0.824)< 0.001SubgroupanalysisBase Model0.655 (0.569–0.740)referenceCL-adjusted Model0.700 (0.617–0.783)0.045 (-0.006-0.097)0.0860.031 (0.006–0.056)0.0160.419 (0.116–0.722)0.008The differences in the area under the receiver operating characteristic curve between the two models were compared using the method of DeLong et al. The incremental predictive performance was evaluated using net reclassification improvement and integrated discrimination improvement. The primary analysis was performed in the overall cohort, while the subgroup analysis included women with a cervical length > 25 mm before 28 weeksAbbreviation: AUC area under the curve, IDI integrated discrimination improvement, NRI net reclassification improvement Assessment of the incremental value of third-trimester short cervix for preterm birth prediction Primary analysis Subgroup analysis The differences in the area under the receiver operating characteristic curve between the two models were compared using the method of DeLong et al. The incremental predictive performance was evaluated using net reclassification improvement and integrated discrimination improvement. The primary analysis was performed in the overall cohort, while the subgroup analysis included women with a cervical length > 25 mm before 28 weeks Abbreviation: AUC area under the curve, IDI integrated discrimination improvement, NRI net reclassification improvement In the subgroup analysis, which included only women with a cervical length > 25 mm in the second trimester, a short cervix status after 28 weeks remained a relevant predictor. The AUC increased from 0.655 (95% CI, 0.569–0.740) to 0.700 (95% CI, 0.617–0.783), although this difference did not reach statistical significance (P = 0.086). However, the CL-adjusted model demonstrated significant improvement in risk reclassification, with an NRI of 0.031 (95% CI, 0.006–0.056; P = 0.016) and an IDI of 0.419 (95% CI, 0.116–0.722; P = 0.008). Discussion In this study, 102 of the 291 included twin pregnancies (35%) resulted in PTB before 37 weeks of gestation. A short cervical length in the early third trimester was associated with a 3.7-fold increased risk of PTB compared to the risk in those without cervical shortening. Furthermore, among women with a normal cervical length in the second trimester, those who subsequently developed cervical shortening in the early third trimester had a 2.3-fold increased risk of PTB. These findings suggest that cervical shortening in the early third trimester is an important marker of increased risk of PTB and may contribute to risk stratification in twin pregnancies. Extensive research has established that twin pregnancies carry a significantly greater risk of PTB than singleton pregnancies [17–19]. Additionally, some studies have reported that cervical length tends to be shorter in twin pregnancies compared to singleton pregnancies [20, 21]. However, recent investigations suggest that, before 28 weeks, cervical length does not significantly differ between singleton and twin pregnancies, whereas, after 28 weeks, cervical length is significantly shorter in twin pregnancies [22]. These observations raise the possibility that cervical shortening observed later in gestation in twin pregnancies may partly reflect physiological changes associated with advancing gestation rather than preterm birth. Nevertheless, such physiological cervical shortening may still be associated with an increased risk of PTB. Accordingly, cervical length measurements obtained in the early third trimester should be interpreted within an appropriate clinical context, as physiological changes may begin to occur while clinically meaningful risk of PTB may still be present. In this study, we found that a short cervical length in the early third trimester was associated with a 3.7-fold increased risk of PTB. This finding is consistent with previous studies reporting the prognostic value of cervical length measured in the late second to early third trimester. Goldenberg et al. reported that a cervical length ≤ 25 mm at 28 weeks was a significant predictor of PTB in twin gestations [23]. However, their analysis evaluated cervical length in combination with other predictors, including fetal fibronectin and bacterial vaginosis assessed between 24 and 28 weeks of gestation, whereas our study focused specifically on cervical length measured in the early third trimester as an isolated risk marker. Similarly, Vayssière et al. reported that cervical length measured at 27 weeks was predictive of spontaneous preterm birth in twin pregnancies, identifying 25 mm as the optimal cutoff based on receiver operating characteristic analysis [24]. Although previous studies have reported an association between cervical length measured in the late second to early third trimester and preterm birth in twin pregnancies, many of these studies evaluated cervical length in combination with other clinical or sonographic parameters, rather than as an isolated predictor. To further assess robustness, analyses using a more stringent cervical length cutoff (≤ 20 mm) yielded results consistent with the primary analysis (Supplementary Tables 4 and 5), suggesting that the association between short cervical length and preterm birth was not dependent on the cutoff definition. Our study extends previous research on the prognostic value of cervical length in the late second to early third trimester by distinguishing between a primary analysis focused on early third-trimester cervical length and a secondary subgroup analysis restricted to women with serial measurements in both the second and early third trimesters. This approach allowed exploratory assessment of both persistent cervical shortening and newly developed shortening in the early third trimester. In a subgroup analysis limited to women with a normal cervical length (> 25 mm) in the second trimester, those who subsequently developed a short cervical length (≤ 25 mm) in the early third trimester had a 2.3-fold increased risk of PTB. This finding aligns with a previous report suggesting that cervical shortening identified for the first time in the early third trimester is associated with an elevated risk of preterm delivery. Park et al. similarly reported that a newly detected short cervix during the third trimester was associated with an 8.61-fold higher risk of PTB before 34 weeks using a cutoff of < 20 mm [15]. Despite differences in cervical length thresholds and outcome definitions across studies, late-onset cervical shortening in the early third trimester appears to be consistently associated with preterm birth in twin pregnancies. In asymptomatic twin pregnancies, cervical length measurement in the early third trimester may provide additional information for risk assessment of PTB. Short cervix status in the early third trimester modestly improved the predictive performance of our model, increasing the AUC from 0.637 to 0.716 (P = 0.001). In addition, improvements in IDI (0.071, P < 0.001) and NRI (0.594, P < 0.001) further indicated enhancement in risk discrimination and stratification for preterm birth. In the subgroup of women in whom cervical shortening was first identified after 28 weeks, the increase in AUC from 0.655 to 0.700 did not reach statistical significance (P = 0.086); however, both IDI and NRI showed consistent trends toward better performance. Taken together, these results suggest that cervical length assessment in the early third trimester may improve risk stratification for PTB, providing complementary information when considered alongside established risk factors in asymptomatic twin pregnancies. However, these findings should be interpreted in the context of the moderate discriminative performance of the model. Kaplan–Meier analysis demonstrated a significantly higher cumulative incidence of preterm birth among women with a short cervix in the early third trimester (Fig. 2). Despite this increased cumulative risk, longitudinal cervical length trajectories did not differ between groups, indicating that the observed differences are more consistent with baseline differences in cervical length than with differences in the rate of cervical shortening over time (Fig. 3). Clinically, evidence regarding the significance of a short cervical length in the early third trimester of twin pregnancies remains limited. Moreover, differences between twin and singleton pregnancies have often made counseling and preventive strategies challenging. In this context, our findings may contribute to the understanding of early third-trimester cervical shortening in twin pregnancies as a potential marker associated with an increased risk of preterm birth, rather than as an indication for specific clinical interventions. Previous studies have reported that continuous cervical length monitoring in twin pregnancies increases antenatal corticosteroid use among those delivering preterm; however, whether such practices translate into improved perinatal outcomes has not been clearly established [21, 25, 26]. Although a recent study suggested that cervical cerclage performed before 24 weeks may reduce PTB in twin pregnancies, its benefit when performed between 24 and 28 weeks is not well defined [27]. Importantly, evidence guiding the management of asymptomatic cervical shortening detected after 28 weeks is currently lacking, and no consensus exists regarding optimal surveillance or intervention strategies in this setting. Nevertheless, early identification of patients with a higher likelihood of PTB may support counseling and general perinatal preparedness, including consideration of referral to appropriately equipped facilities, although its impact on clinical decision-making requires further prospective validation. Accordingly, further research is needed to determine how information derived from early third-trimester cervical length assessment can be most effectively integrated into clinical care for twin pregnancies. This includes assessing the effectiveness of interventions such as vaginal progesterone, surveillance strategies, or other clinical actions specific to this later gestational window. This study has several strengths. First, it used a large dataset collected over a 10-year period at a single tertiary care center, allowing for standardized cervical length assessment within a relatively uniform clinical setting. Second, the study highlights the potential value of longitudinal cervical length monitoring, rather than reliance on a single mid-trimester measurement. The use of least squares means plots facilitates visualization of temporal patterns in cervical length across gestation and improves interpretability of longitudinal differences between the preterm birth and control groups. Importantly, by focusing on cervical length assessment in the early third trimester and distinguishing newly developed shortening from earlier measurements, this study extends prior literature that primarily emphasized mid-trimester assessments. Third, the inclusion of complementary performance metrics, including the IDI and NRI, provides additional support for the incremental predictive value of early third-trimester cervical length in risk stratification models. However, this study also has certain limitations. First, as a retrospective study, there is potential for selection bias, and missing data may have influenced the results. A substantial proportion of twin pregnancies were excluded during the selection process due to prior cervical cerclage, iatrogenic preterm birth, or the absence of cervical length measurements suitable for analysis. Given the tertiary referral nature of our institution, many excluded cases presented with advanced cervical dilatation or active preterm labor and delivered shortly after admission, precluding cervical length assessment. To evaluate the impact of this exclusion, we compared maternal characteristics, factors related to cervical length assessment, and delivery outcomes between the included and excluded pregnancies (Supplementary Table 1). The excluded population demonstrated significantly earlier gestational age at delivery and higher rates of preterm birth, indicating that excluded cases represented a higher-risk group. Accordingly, higher-risk twin pregnancies, such as those with active preterm labor, advanced cervical dilatation, or those requiring cervical cerclage, were likely underrepresented in the analytic cohort. Consequently, the study population may not fully reflect the broader twin pregnancy population, and this underrepresentation should be considered when interpreting the generalizability of our findings. Second, some additional risk factors for PTB, including maternal infections, inflammatory markers, and twin-specific complications such as twin-to-twin transfusion syndrome, selective fetal growth restriction, and twin anemia–polycythemia sequence, may not have been fully accounted for in the analysis. Third, clinical interventions initiated after identifying a shortened cervical length, such as bed rest or intravenous fluid administration, may have affected the study outcomes. In addition, although progesterone use was adjusted for in the analysis, residual confounding by indication cannot be fully excluded, as treatment may have been preferentially administered to patients at higher baseline risk. Despite these limitations, this study suggests that early third-trimester cervical length measurements may offer additional value in risk stratification of preterm birth in twin pregnancies. Future prospective, multicenter studies are needed to validate these findings and refine strategies for risk stratification and perinatal management in high-risk twin pregnancies. Conclusions This study indicates that a short cervical length (≤ 25 mm) detected in the early third trimester is significantly associated with an increased risk of PTB in twin pregnancies. Moreover, even among women with a normal cervical length in the second trimester, those whose cervical length became shorter in the early third trimester also had a higher risk of PTB. These findings suggest that early third-trimester cervical length assessment may serve as a valuable tool for evaluating PTB risk and optimizing perinatal management in twin pregnancies, particularly among asymptomatic or relatively lower-risk populations. Supplementary Information Supplementary Material 1. Acknowledgements We would like to express our sincere gratitude to all the physicians and sonographers who contributed to data collection. Abbreviations Preterm birth In vitro fertilization Loop electrosurgical excision procedure Interquartile range Area under the receiver operating characteristic curve Net reclassification improvement Integrated discrimination improvement Authors' contributions OK: Writing – original draft, methodology, investigation, formal analysis, and visualization HYL: Data curation, Resources HSK: Data curation, funding acquisition, and investigation IYP: Data curation, and investigation, Resources, and Validation SH: Writing – review and editing, conceptualization, formal analysis, funding acquisition, methodology, supervision, and project administration All authors reviewed and approved the final manuscript. Funding This research was supported by a grant of Patient-Centered Clinical Research Coordinating Center (PACEN) funded by the Ministry of Health & Welfare, Republic of Korea (grant number: RS-2025-25410838). Data availability The datasets are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate This study was approved by the Institutional Review Board (IRB) of Seoul St. Mary’s Hospital (IRB no. KC23RISI0613) and conducted following the ethical standards established in the Declaration of Helsinki. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The datasets are available from the corresponding author on reasonable request.