Diagnostic performance and feasibility of classifying placental umbilical cord insertion during routine prenatal ultrasound

Abstract

Introduction

Abnormal cord insertion (velamentous or marginal) is associated with adverse perinatal outcomes, and guidelines recommend prenatal diagnosis when feasible. However, the diagnostic performance and feasibility of classification of the placental cord insertion in an ultrasound screening setting remain uncertain. The aim of the study was to evaluate the feasibility and diagnostic performance of prenatal classification of placental cord insertion site.

Material and Methods

This prospective observational study was conducted at a Norwegian tertiary center from May 2017 to August 2023. Pregnant women with singleton pregnancies attending the routine ultrasound scans were invited to participate and enrolled after written informed consent. During the scan, the distance between the umbilical cord insertion and the placental edge (the “test variable”) was measured. At birth, cord insertion was categorized as normal, abnormal (velamentous (≤0 mm) or marginal (1–24 mm) from the placental edge), or unclassifiable. Diagnostic performance of the measured distance by ultrasound to predict insertion type at birth was evaluated using area under the receiver operating characteristic curve (AUC), with adjustment for covariates in Bayesian logistic regression, including sonographer experience. Sensitivity, specificity, and positive and negative predictive values, including optimal thresholds (Youden and Closest top‐left criteria), were estimated.

Results

In total, 2859 examinations were carried out, and 2553 examinations in 1919 pregnancies were between 11 and 18 gestational weeks and included in the analyses. Of these, 634 had two examinations. Cord insertion site was classifiable in all but 14/2859 examinations (0.49%). The AUC for prenatal identification of abnormal cord insertion was 0.76. Using the Youden threshold, sensitivity was 0.64 and specificity 0.76; using the Closest top‐left threshold, sensitivity was 0.69 and specificity 0.70. Negative predictive values exceeded 0.90, whereas positive predictive values (PPVs) were lower. Diagnostic performance was higher for velamentous cord insertion specifically (AUC = 0.89) than for abnormal insertion overall. Adjusting for additional covariates (operator experience and maternal characteristics) minimally affected AUC estimates.

Conclusions

Assessment of the umbilical cord insertion during routine prenatal ultrasound was feasible and demonstrated high negative but limited PPV, suggesting that cord insertion assessment may be incorporated into routine antenatal imaging to support risk stratification.

Classification of placental umbilical cord insertion is feasible within the routine prenatal ultrasound examination, and reliably excludes abnormal insertion, although positive predictive values were low.

Abbreviations

AUC

area under the receiver operating characteristic curve

CI

confidence interval

MCI

marginal cord insertion

NPV

negative predictive value

PPV

positive predictive value

ROC

receiver operating characteristic

VCI

velamentous cord insertion

Key message.

Routine prenatal ultrasound allows feasible classification of placental umbilical cord insertion and reliably excludes abnormal insertion, although positive predictive values were low.

1. INTRODUCTION

Abnormal cord insertion includes two main entities: velamentous and marginal insertion (VCI and MCI). In VCI, the umbilical cord inserts into the fetal membranes rather than the chorionic plate, while in MCI, the insertion is on the placenta but close to the edge. A variant of velamentous vessels also occurs in bilobed placentas, where fetal vessels run through the membranes connecting the lobes. In both VCI and this variant, fetal vessels traverse unprotected in the membranes.

VCI (occurrence 1.5%) ¹ is associated with an increased risk of adverse outcomes for both mother and neonate. ¹ , ² , ³ , ⁴ These risks are attributed to abnormal placental development, ⁵ potential obstruction of fetal flow and hemorrhage, which may occur from velamentous vessels after rupture of the membranes before or during delivery. ⁶ Early detection allows individualized management during pregnancy and birth, particularly important in cases with vasa previa (VP), the variant of VCI where the fetal vessels run in the membranes near the internal cervical os. If undiagnosed, VP is associated with a high risk of perinatal mortality and morbidity, while with a prenatal diagnosis the neonatal outcome is significantly improved. ⁷ , ⁸ , ⁹

Previous studies have shown that prenatal ultrasound, including color Doppler, can classify both MCI and VCI, ¹⁰ , ¹¹ but the feasibility of incorporating this assessment into population‐based antenatal screening remains unknown. In Norway, antenatal care is publicly funded, and most pregnant women attend routine ultrasound scans ¹² that are almost exclusively carried out by specially trained midwives with uniform education. ¹³ Current Norwegian guidelines do not mandate assessment of the placental cord insertion during the routine scans. ¹⁴ Before universal implementation, it is essential to evaluate feasibility and diagnostic performance and to identify factors that may influence diagnostic performance, such as the examiner's level of experience, gestational age, maternal body mass index (BMI), and placental position.

The aim of the study was to assess the feasibility and diagnostic performance of classifying the placental cord insertion site during the routine ultrasound screening in Norway.

2. MATERIAL AND METHODS

This prospective observational study was conducted at the Fetal Medicine Unit, Haukeland University Hospital, from May 24th, 2017 – August 5th, 2023. Women with singleton pregnancies attending the routine scan were invited and included after written consent. Depending on gestational age, participants could contribute more than one examination.

The primary outcome was the area under the curve (AUC) for diagnostic performance. The main exposure was the shortest distance as a continuous variable from the umbilical cord insertion to the placental edge assessed by ultrasound.

During the routine 30‐min transabdominal ultrasound examination, the umbilical cord insertion was identified using grayscale and color Doppler imaging (Figure 1A). The shortest distance from the cord insertion to the placental edge was measured and used to classify the insertion as normal (≥25 mm), marginal cord insertion (MCI; 1–24 mm), or velamentous cord insertion (VCI; insertion into the membranes, defined as ≤0 mm). ¹⁵ Insertions that could not be reliably assessed were classified as unclassifiable. Cases with abnormal or unclassifiable cord insertion were offered follow‐up scans by a physician. Midwives followed a standardized protocol, assessing placental location, maximal placental diameter and shortest cord‐to‐edge distance using track‐distance measurements (Figure 1B). Examiners received a single theoretical introduction at study start and were blinded to prior measurements for pregnancies with repeated scans.

FIGURE 1.

(A) Ultrasound image of the placenta showing the umbilical cord insertion (arrow), identified using color Doppler technique. (B) Ultrasound image showing the shortest distance from the outer edge of the umbilical cord insertion to the placental edge by a linear trace measurement (red line).

If VCI was suspected, the women were offered transvaginal ultrasound at 20–24 weeks to exclude VP and a third‐trimester growth assessment in the clinical routine, but these examinations were not in the study protocol. After delivery, attending midwives recorded cord insertion (normal, MCI, VCI) in the obstetric patient record (Natus®).

Demographic and pregnancy data for both study participants and all singleton births in our department during the study period were retrieved from Natus®. Postnatal classification defined VCI as insertion into the membranes (≤0 mm) (Figure 2), and MCI as insertion within 25 mm from the placental edge.

FIGURE 2.

Photo of the placental fetal side with umbilical cord attached in the membranes, (VCI) and with velamentous vessels. (Measuring tape indicates the placental diameter.)

Feasibility was defined as successful visualization and classification of the cord insertion within the routine 30‐min scan. Thirteen midwives performed scans. Their level of experience was categorized (4 levels) (Table S1) and included in receiver operating characteristic (ROC) models as covariates. Further, maternal BMI, placental position, and gestational age at examination were evaluated as factors potentially influencing diagnostic performance.

The ultrasound systems used were Voluson E6® and GE Voluson E8® equipped with abdominal and endocavity transducers.

2.1. Statistical analyses

We estimated performance metrics (sensitivity, specificity, PPV, and NPV) of ultrasound‐measured distance from the umbilical cord insertion to the placental edge to determine abnormal cord insertion at birth. Bayesian logistic regression was used to model the relationship between distance from the umbilical cord insertion to the placental edge as exposure (“test variable”) and abnormal cord insertion at birth, generating ROC curves and ROC tables, and calculating AUC. Bayesian regression was chosen to efficiently utilize the data. We used density plots and Markov Chain Monte Carlo (MCMC) trace plots of regression parameters to ascertain convergence of the models (Appendix S1). To stabilize computation, we used default weakly informative priors. ¹⁶ The diagnostic performance of a ROC curve based on log concave density (distribution) of the exposure is usually better than that of an empirical ROC curve. ¹⁷ We therefore assessed if the exposure variable (distance) exhibited a log concave density using bootstrap p‐values.

Because the measured distance varied with gestational age, distances were standardized to a common gestational age (18 complete weeks) prior to analysis, using a natural spline regression model (R‐package splines) ¹⁸ including a single or all available measurements. For pregnancies with two scans, we considered including the first or second measurement, or the two combined, and used the alternative with the highest AUC in the final analyses. ¹⁹

Predicted “posterior” probabilities obtained from the Bayesian regression models were used to estimate sensitivities, specificities, PPV, and NPV including “optimal” thresholds (Youden and Closest top‐left). For clinical interpretability the metrics were aligned with specific distance thresholds (Tables 2, S2 and S3). The 95% CIs around the ROC curves were estimated using 10⁴ nonparametric bootstrap replicates. The 95% credible intervals around the metrics were calculated using Bayesian standard deviations (Table S3). Covariates (gestational age at examination, maternal BMI, maternal age, parity, placental position (previa, anterior, posterior, fundus, left and right lateral), operator experience and year of examination) were added to the Bayesian model to assess their potential contribution to the AUC. We also calculated performance metrics from absolute numbers and compared them with those based on Bayesian regression (Table S2).

TABLE 2.

Diagnostic performance across cutoff values of the distance between the placental edge to the umbilical cord insertion, measured during routine scans (11–18 weeks).

Cutoff (mm)	Sensitivity	Specificity	PPV	NPV	Best threshold a
Abnormal placental cord insertion at birth
<30	0.78	0.58	0.19	0.96
<26.5	0.69	0.70	0.22	0.95	Closest top‐left b
<25	0.65	0.74	0.23	0.95
<24.5	0.64	0.76	0.24	0.94	Youden c
<15	0.33	0.95	0.45	0.92
≤0	0.11	1.00	0.78	0.90
Diagnostic performance for velamentous cord insertion at birth
<30	0.91	0.55	0.04	1.00
<25	0.86	0.72	0.06	1.00
<19.5	0.79	0.85	0.10	1.00	Closest top‐left b
<16.5	0.74	0.91	0.15	0.99	Youden c
<10	0.64	0.97	0.34	0.99
≤0	0.50	1.00	0.71	0.99

Note: ROC table with sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for different cutoff values (mm) of the distance. Data are based on 2553 examinations in 1919 women, and the metrics were estimated using Bayesian logistic regression.

^a Robin et al. ²³

^b Closest top‐left coordinates.

^c Youden coordinates.

Sample size estimation was performed a priori using a nomogram ²⁰ and confirmed during the study using the R‐package epiR. ²¹ With a sensitivity to detect abnormal insertion of the umbilical cord of 0.95, an incidence of velamentous or marginal umbilical cord insertions of 0.075 at birth and a 95% confidence interval of the result, the estimated study size was 974 pregnancies for the first and second trimester routine scans, respectively. We increased the sample size to >2000 pregnancies to account for dropouts, and for the possibility to study the influence of gestational age at the examinations on feasibility and diagnostic performance.

IBM SPSS Statistics (Version 29.0), Stata (version 18.5), and R packages epiR version 2.0.88, ²¹ rstanarm version 2.32.2, ¹⁶ pROC version 1.19.0.1, ²² and splines version 4.5.1 ¹⁸ were used for statistical analysis.

3. RESULTS

A total of 2129 pregnancies contributed with 2859 observations. Of these, 2553 in 1919 pregnancies were performed at 11–18 weeks and included in analyses. Of these, 634 women had more than one scan. Characteristics of the study and background populations are given in Table 1. Women in the study sample more frequently had normal BMI (18.5–24.9 kg/m²) than in the background population (60.1% vs. 58.5%, respectively), and data on BMI were more often missing (3.0%, vs. 1.7%). Conception by in vitro fertilization (IVF) was more frequent in the study sample (9.0% vs. 5.6%). A lower proportion of participants were of non‐Norwegian origin compared with the background population (Table 1).

TABLE 1.

Characteristics of the study population (n = 1919) and reference population (N = 27 486) of singletons giving birth in our population during the study period.

	n (study population)	%	N (reference population)	%
Parity
0	818	42.6	12 198	44.4
1	703	36.6	10 384	37.8
2+	355	18.5	4899	17.8
Missing	1	0.1	5	0.0
Maternal age
<20	6	0.3	90	0.3
20–24	115	6.0	1699	6.2
25–29	463	24.1	7616	27.7
30–34	604	31.5	11 315	41.2
35–39	545	28.4	5404	19.7
40+	186	9.7	1359	4.9
Missing	0	0.0	3	0.0
Maternal BMI (kg/m2)
<18.5	53	2.8	868	3.2
18.5–24.9	1154	60.1	16 073	58.5
25–29.9	417	21.7	6330	23.0
30–34.9	140	7.3	2509	9.1
35+	55	2.9	1236	4.5
Missing	58	3.0	470	1.7
Mothers' origin
Norway	1677	87.4	22 959	83.5
Western	99	5.2	1842	6.7
Non‐Western	101	5.3	2680	9.8
Missing	0	0.0	5	0.0
Conception by IVF	172	9.0	1546	5.6
Velamentous insertion	39	2.0	658	2.4
Marginal insertion	170	8.9	2446	8.9
Diabetes Mellitus type I	10	0.5	169	0.6
Diabetes Mellitus type II	3	0.2	58	0.2
Chronic Hypertension	9	0.5	140	0.5
Sex of the neonate Boy	930	48.5	14 118	51.4
Girl	946	49.3	13 356	48.6
Missing	1	0.0	12	0.0
Apgar score <7 (5 min)	33	1.7	666	2.4
Delivery mode
Vaginal birth	1674	87.2	21 936	79.8
Elective caesarean section	56	2.9	809	2.9
Transferred to NICU	151	7.9	2497	9.1
Gestational age at birth
Preterm	79	4.1	1695	6.2
Term	1611	83.9	23 083	84.0
Post term	187	9.7	2708	9.9

Note: Preterm <259 days, Term 259–291, Postterm >291 days. 42 births were in other hospitals. Data from these births are missing.

Abbreviations: IVF, in vitro fertilization; NICU, neonatal intensive care unit.

The placental cord insertion site was classified within the allocated time in all but 14 examinations, 14/2859 (0.49%). Of the 1919 pregnancies included, 170 and 39 were diagnosed with marginal and VCI at birth, respectively (Table 1). The distribution of distances between cord insertion and placental edge demonstrated log concave density (bootstrap p > 0.05) after exclusion of 10 extreme values, and the ROC analyses were therefore based on this assumption.

For pregnancies with two scans, the statistical function combining repeated measures provided the highest AUC, but the differences between the approaches were small (<0.02). For abnormal cord insertion overall, the crude AUC was 0.76 (95% CI: 0.73–0.80) (Figure 3). Density plots and MCMC trace plots for the regression parameters indicated model convergence (Appendix S1).

FIGURE 3.

Receiver operation characteristic (ROC) curve with 95% confidence band for prediction of abnormal umbilical cord insertion at birth from for the shortest distance between the placental edge and the umbilical cord insertion during the first and second trimester routine ultrasound examination.

For the diagnosis of abnormal cord insertion, sensitivity increased with the distance from the umbilical cord insertion to the placental edge (Tables 2, S2 and S3). With Youden thresholds sensitivity was 0.64 and specificity 0.76, corresponding to distance <24.5 mm. With the Closest top‐left threshold (distance <26.5 mm) sensitivity was 0.69 and specificity 0.70. Adapting the postnatal criterion of <25 mm to the prenatal situation, sensitivity and specificity were intermediate to those of the Youden and Closest top‐left cutoffs (specificity and sensitivity 0.65 and 0.74). NPVs were above 0.90 while PPV's were lower.

Using a threshold of <25 mm, the corresponding false‐positive rate for abnormal cord insertion was 24% (1 – specificity = 0.76, Table 2), reflecting the proportion of pregnancies with misclassified abnormal cord insertion.

For postnatal VCI, the crude AUC was 0.89 (0.86–0.94) (Figure 4). For VCI, sensitivity increased with the distance between the placental edge and the umbilical cord insertion (Tables 2, S2 and S3). With Youden thresholds sensitivity was 0.74 and specificity 0.91, corresponding to <16.5 mm. With the Closest top‐left threshold (distance <19.5 mm) sensitivity was 0.79 and specificity 0.85. NPV was 0.99–1.00 while PPV was low, except for distance ≤0 mm (prenatal VCI).

FIGURE 4.

Receiver operation characteristic (ROC) curve with 95% confidence band for prediction of velamentous umbilical cord insertion at birth from for the shortest distance between the placental edge and the umbilical cord insertion during the first and second trimester routine ultrasound examination.

Adding covariates, including examiner experience, placental position, examination year, maternal BMI, maternal age, parity, and gestational age, changed AUC estimates by <0.01 for both abnormal insertions overall and VCI.

Diagnostic performance estimates using Bayesian regression agreed with those calculated from absolute numbers but were different for thresholds with small numbers (Table S2).

4. DISCUSSION

We found that incorporating visualization of the placental cord insertion site into ultrasound screening was feasible, and that the antenatal classification shows acceptable diagnostic performance, which was best for VCIs.

Specificity and NPV were high (Table 2), supporting integration of the cord insertion assessment into the standard protocol. The modest sensitivities indicate that some cases will be missed. As in prior work, ²⁴ PPVs were low, particularly for VCI, partly reflecting a larger denominator with false‐positives (Table S2). Physiological regression of MCI may explain some of these false alarms. ²⁵ Accordingly, counseling should stress diagnostic uncertainty.

Adding covariates including placental position resulted in minimal changes in AUC (<0.01), which suggests that the diagnostic performance of the method is relatively robust across varying clinical and operator characteristics.

Although our study population had a slightly more favorable BMI, potentially improving visualization, maternal BMI did not significantly affect AUC. While VCI can be identified as early as during gestational week 11–14 weeks, ¹⁰ our data did not show materially different performance by gestational age. The high AUC for VCI in our study suggests a relative biological stability of this condition in contrast to marginal insertions, consistent with the report from Ward et al. reporting that the placental cord insertion has the potential to evolve due to heterogeneous growth of the placenta. ²⁵

Strengths of the study include the prospective design, the real‐world setting, its adequate size, and postnatal classification. The study was carried out within the first and second trimester routine examinations without added time. The different levels of experience of the 13 examining midwives enhance the external validity. Conversely, the reliability of the diagnostic performance estimates could be reduced due to the relatively low numbers of the VCI births (n = 39, Table 1). However, using Bayesian regression analysis improves estimation for rare events and yields more reliable point‐ and interval estimates than with a frequentist (“classical”) approach. ²⁶ Attending midwives at delivery were blinded to prenatal cord insertion assessments (except suspected VCI), reducing classification bias postpartum; prior work in our institution showed good interobserver agreement for postnatal placental and cord classifications. ²⁷

A substantial proportion of the pregnancies provided data from more than one examination. While integrating repeated distances between the cord insertion and placental margin may be cumbersome and less clinically intuitive to interpret than using a single measurement, disregarding pertinent data remains less methodologically appropriate. ¹⁹

We included more participants than the suggested number from the calculation of sufficient study size to consider potential dropouts, the unknown success rate for identification of the cord insertion, and to get a sufficient distribution of examinations over the gestational age period. Despite the universal and no‐cost access to prenatal screening in Norway, exclusion of non‐Norwegian‐speaking women may limit generalizability. ²⁸ In addition, during the early study years and before universal implementation in 2022, first‐trimester scans were indication‐based, potentially increasing the number of high‐risk pregnancies in our cohort, also reflected in the higher proportion of pregnancies conceived by IVF in the study group (Table 1).

A methodological original contribution is the ROC analysis, based on antenatally measured distance between the placental edge and the cord insertion site. Although performance characteristics of identifying abnormal cord insertion prenatally have been reported, ²⁴ , ²⁹ , ³⁰ , ³¹ these analyses were based on one fixed distance, varying between the studies (1–2 cm). Consequently, ROC analyses could not be performed, and these studies could not be easily compared.

Although the study was conducted in a tertiary center, the midwives had no study‐specific training in cord insertion classification. Given a common national curriculum, ¹³ , ³² our findings are likely generalizable to other settings. Cord insertion assessment was rapidly adopted by the midwives irrespective of experience, and adding experience to the ROC model did not alter performance, in contrast to reports linking sonographers' experience with visualization rates of congenital heart defects. ³³ The European Association of Perinatal Medicine recommends assessing the cord insertion as part of the routine prenatal imaging, ³⁴ while it is only included for risk pregnancies in the American Institute of Ultrasound in Medicine and ISUOG Practice guideline recommendations. ³⁵ , ³⁶

Our feasibility data show that cord insertion can be visualized and distance to edge measured within the frame of the screening examination, supporting integration into the 11–18 weeks scans without significant burden. However, feasibility as defined in our study does not capture diagnostic accuracy. This is addressed in the analyses of the diagnostic performance (AUC), which was overall good, with consistently high NPV indicating that the method is reliable for ruling out abnormal cord insertion.

The false‐positive rate is dependent on the selected threshold, illustrating the expected trade‐off between sensitivity and specificity. A stricter cutoff will result in fewer false‐positives and consequently less follow‐up examinations. In clinical practice, false‐positive findings would prompt targeted ultrasound examinations, where the benefit of detection likely would outweigh the burden of the additional examinations.

Cord insertion classification is safe and cost‐neutral, features that are essential for any test considered for use in a screening program. ³⁷ However, abnormal findings can provoke anxiety and balanced counseling remains essential. Downstream management and cost‐effectiveness of universal vs. targeted cord insertion screening are not yet established. ¹¹ , ³⁸ For VP targeted screening based on ultrasound findings or risk factors such as pregnancies after assisted reproductive technology, it appears more cost‐effective than universal screening and falls within acceptable willingness‐to‐pay thresholds in a study from the USA. ³⁹ Two‐stage strategies, consisting of universal assessment of placenta and cord insertion at the routine scan followed by a selection of patients to targeted transvaginal ultrasound including color Doppler to rule in or out VP, are feasible and may reduce the rate of stillbirths by up to 10%. ⁴⁰ The UK National Screening Committee nonetheless judged first‐trimester VP screening to have limited benefit due to the placental evolution. ⁴¹

Further work should evaluate the cost‐effectiveness of universal versus targeted cord insertion screening and clarify how placental growth dynamics shape placental form and cord insertion development.

5. CONCLUSION

Assessment of the umbilical cord insertion during routine prenatal ultrasound was feasible and demonstrated high negative, but limited positive predictive value, suggesting that cord insertion assessment may be incorporated into routine antenatal imaging to support risk stratification.

AUTHOR CONTRIBUTIONS

Ine Hildershavn Moen: Conceptualization; methodology; project administration; investigation; data curation; writing – original draft; writing – review and editing. Synnøve Lian Johnsen: Review and editing. Jørg Kessler: Data curation; review and editing. Alexander Vietheer: Data curation; review and editing. Svein Rasmussen: Conceptualization; methodology; supervision; formal analysis; writing – original draft; writing – review and editing. Cathrine Ebbing: Conceptualization; methodology; project administration; writing – original draft; writing – review and editing. All authors approved the final manuscript and agreed to be accountable for all aspects of the work.

FUNDING INFORMATION

No additional funding was obtained for the conduct of this study.

CONFLICT OF INTEREST STATEMENT

None declared.

ETHICS STATEMENT

The study was approved by the Regional Ethics Committee (REC West 2011/949) on December 21, 2016.

Supporting information

Appendix S1: Visualization of model convergence using Markov Chain Monte Carlo (MCMC) in Bayesian models. Panel A: Outcome is postpartum abnormal cord insertion. Panel B: Outcome is postpartum velamentous cord insertion. The plots on the left display the results (spread around the mean) of four independent simulation series (chains) for the regression coefficients (intercept and the distance between the placental margin and the umbilical cord insertion on the logit scale, first measurement). The four density (spread) plots are essentially congruent. The trace plots on the right illustrate the progression of the simulated regression coefficients across increasing iterations. Following an initial 3000 “warmup” iterations, an additional 10 000 iterations were run and retained. The trace plots exhibit characteristic “hairy caterpillar” patterns, indicating that the MCMC chains have run long enough to provide valid results (i.e., the models have converged to a stationary distribution).

Table S1. The different levels of experience of the 13 examining midwives (4 levels).

Table S2. Numbers of abnormal (ACI) and velamentous cord insertion (VCI) at birth by cutoff points of distance from the placental edge to the umbilical cord insertion at examination. Diagnostic metrics were based on absolute numbers or Bayesian regression. Around 1919 pregnancies (2553 examinations) with gestational age 11–18 weeks.

Table S3. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) with 95% Bayesian credible intervals (CrI) by cutoff values (mm) of the distance from the placental edge to the umbilical cord insertion, measured during routine scans (11–18 weeks). Around 1919 women (2553 examinations).

Moen IH, Johnsen SL, Kessler J, Vietheer A, Rasmussen S, Ebbing C. Diagnostic performance and feasibility of classifying placental umbilical cord insertion during routine prenatal ultrasound. Acta Obstet Gynecol Scand. 2026;00:1‐9. doi: 10.1111/aogs.70359

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.