An algorithm for predicting Robin sequence from fetal MRI

Background: Infants with Robin sequence (RS) may present with airway compromise at delivery. Prenatal diagnosis would improve preparation and postnatal care. The purpose of this study was to devise a predictive algorithm for RS based on fetal magnetic resonance imaging (MRI).
Methods: Retrospective case-control study including fetal MRIs from 2002 to 2017. Inclusion criteria were (1) MRI of adequate quality, (2) live-born infant, and (3) postnatal evaluation. Subjects were grouped on the basis of postnatal diagnosis:
(1) RS (micrognathia, glossoptosis, airway obstruction),
(2) micrognathia without airway obstruction ("micrognathia"),
(3) cleft lip and palate ("CLP"), and
(4) gestational age-matched controls.
A series of possible predictive variables were assessed on MRI. Receiver operator curves were applied to identify cut-off values, and a multivariable algorithm was developed.
Results: A total of 162 subjects with mean gestational age at MRI of 25.6 ± 4.9 weeks were included: RS, n = 27 (17%); micrognathia, n = 35 (22%); CLP, n = 46 (28%); control, n = 54 (33%). Three variables were independent predictors of RS: (1) Veau I/II cleft palate (OR = 38.8), (2) tongue shape index (>80%; OR = 8.7), and (3) inferior facial angle (<48° OR = 14.5).
Conclusion: MRI findings of cleft palate, TSI >80% and IFA <48° indicate a 98% probability of RS, whereas a lack of all 3 features denotes a likelihood of 1%.