Introduction: Prenatal counseling regarding atrioventricular septal defect (AVSD) surgical outcomes in patients with borderline small left heart structures remains a challenge. Our aim was to identify fetal echocardiographic variables associated with single ventricle (SV) palliation, staged biventricular (BiV) repair, or standard BiV repair. 

Methods: A retrospective cohort study was performed including fetuses with AVSD (2011-2022). Those with valve atresia, heterotaxy, great artery malposition, trisomy 13/18, diaphragmatic hernia, growth restriction, fetal death, comfort care, or missing follow-up were excluded. Fetal echocardiographic measurements with z-scores were collected. Aortic-valve-to-pulmonary-valve annulus ratios and short-axis left-to-right-ventricular end-diastolic dimension (LVED-RVED) ratios were calculated. Outcomes included SV palliation, staged BiV repair, or standard BiV repair. First, univariable mixed regression analysis was performed to evaluate associations between fetal variables and SV palliation. Then, classification and regression tree (CART) analyses using all echocardiograms were performed to construct an algorithm to predict surgical outcomes. 

Results: Of 98 fetuses (209 echocardiograms) included, Down syndrome was present in 77%.  Six (6%) underwent SV palliation, 14 (14%) underwent staged BiV repair, and 78 (80%) underwent standard BiV repair. No child underwent aortopulmonary shunt or ductal stent placement as part of staged BiV repair or SV palliation. In univariable analysis, lower z-scores of short-axis LVED, long-axis LVED, transverse arch and isthmus were associated with SV palliation, as were qualitative assessments of an unbalanced AVSD or non-apex-forming LV, lower aortic-valve-to-pulmonary-valve or short-axis LVED-RVED ratios. Including all 129 echocardiograms with complete measurements, CART modeling identified LVED-RVED ratio <0.61 with absence of predominantly right-to-left flow at the foramen ovale as predictive of SV palliation (71%), and these measures outperformed qualitative unbalanced assessment. In contrast, LVED-RVED ratio >0.61 and transverse arch z-score >-1.87 had a 96% chance of standard BiV repair, and a 4% chance of staged BiV repair, with no SV palliation. Repeat CART modeling excluding LVED and RVED measurements (171 echocardiograms) identified either 1) transverse arch z-score <-1.88 with aortic valve z-score <-2.77 or 2) qualitatively unbalanced AVSD and left-to-right or bidirectional foramen ovale flow in those with normal arch measurements as predictive of SV palliation (89% and 67% respectively). 

Conclusion: Lower LVED-RVED ratio measurements were associated with SV palliation in prenatally diagnosed fetuses with AVSD. In addition, other surrogate objective markers for degree of unbalance, such as aortic-valve-to-pulmonary-valve annulus ratio, appear promising as predictors of staged BiV repair among well-balanced patients. CART model decision trees remain a powerful tool for improving prenatal anticipatory guidance for congenital heart disease. Future studies should validate CART model cutoffs prospectively.