Sagittal aortic‐arch assessment for prenatal diagnosis of coarctation of the aorta: comparative evaluation of 10 published methods and a new predictive model

ABSTRACT Objectives To externally validate the diagnostic performance of 10 previously published fetal echocardiographic sagittal aortic‐arch parameters and to develop an optimal predictive model based on these parameters for the prenatal identification of neonatal coarctation of the aorta (CoA). Methods This was a retrospective, multicenter diagnostic cohort study, including pregnancies with suspected fetal CoA and postnatal diagnostic confirmation. Sagittal aortic‐arch videoclips obtained from patients at two tertiary referral centers in Thailand between January 2014 and January 2025 were reviewed offline by a single investigator blinded to the final diagnosis to assess 10 predefined echocardiographic parameters of the aortic arch, including vascular diameters, interarterial distances, vessel angles and the presence of a shelf sign. The discriminatory performance of the 10 published predictors for neonatal CoA was evaluated using univariate binary logistic regression and receiver‐operating‐characteristics (ROC)‐curve analysis. A final predictive model was developed by entering the 10 sagittal aortic‐arch parameters into a multivariable logistic regression model and applying backward elimination to derive the model with the fewest parameters without compromising discriminatory performance. Results Of the 154 fetuses included in the study, 82 (53.2%) were subsequently diagnosed with neonatal CoA. Fetuses subsequently diagnosed with neonatal CoA had a significantly longer left common carotid artery–left subclavian artery (LCCA–LSCA) distance (median, 3.12 mm vs 2.14 mm; P = 0.001) and a higher prevalence of the shelf sign (74.4% vs 50.0%; P = 0.002). Among individual parameters, the carotid–subclavian artery index showed the highest diagnostic accuracy for predicting neonatal CoA (area under the ROC curve (AUC), 0.684 (95% CI, 0.600–0.767)). The final multivariable model, incorporating LCCA–LSCA distance, LSCA diameter, descending aortic diameter and the presence of the shelf sign, demonstrated moderate performance (AUC, 0.749 (95% CI, 0.672–0.826)). At a predicted probability cut‐off of 0.45, the sensitivity and specificity of the combined model for the prenatal identification of neonatal CoA were 84.1% (95% CI, 74.4–91.3%) and 54.2% (95% CI, 42.0–66.0%), respectively. Conclusions In fetuses with suspected CoA, a multiparametric predictive model based on sagittal aortic‐arch assessment improves prenatal diagnosis compared with previously published methods. A longer LCCA–LSCA distance, the presence of the shelf sign and an increased LSCA diameter are useful for distinguishing true‐positive cases of CoA during fetal echocardiography. © 2026 International Society of Ultrasound in Obstetrics and Gynecology.