Endorsing consensus interpretation for diagnosing placenta accreta spectrum disorders on MRI.

OBJECTIVES

To determine the performance of MRI features alone and combined in a model for the diagnosis of placenta accreta spectrum (PAS) disorders in pregnant women.

MATERIALS AND METHODS

This single health system IRB-approved retrospective study included 131 pregnant women who underwent MRI for placental (n = 122) or fetal (n = 9, controls) indications (2016-2021). Three independent observers assessed 11 PAS features on MRI, endorsed in prior studies, with consensus adjudication of discordances by 2 separate expert observers. The reference standard was established by intra-operative findings and/or pathology. Univariable and multivariable logistic regression and ROC analysis evaluated MRI feature performance for PAS diagnosis alone and combined in a model.

RESULTS

131 pregnant patients (mean age 36 years, range 21-54 years; mean gestational age 28 weeks) were included. PAS was present in 54/131 (41.2%) patients. Consensus adjudication demonstrated that all MRI features and history of prior deliveries (vaginal or caesarean section) were associated with PAS (p ≤ 0.0003). Most MRI features, except the "serosal vessel sign," were associated with PAS for the independent observers (p < 0.04). At consensus, "loss of T2-retroplacental line" (sensitivity 0.87/specificity 0.43/p = 0.0003) and "myometrial thinning" (sensitivity 0.87/specificity 0.47/p < 0.0001) showed the highest sensitivity for PAS, while "bladder vessel sign" had the highest specificity (sensitivity 0.28/specificity 1.0/p < 0.0001). A multivariable model comprised of "uterine/placental bulge," "focal exophytic placental mass," and "bladder vessel sign" demonstrated an AUC = 0.83 (95% confidence interval 0.76-0.83) for PAS diagnosis using consensus data.

CONCLUSION

A model based on consensus interpretation of MRI findings demonstrated good performance for diagnosing PAS.

KEY POINTS

Question The lack of validated MRI criteria and inter-observer variability limits MRI's reliability for noninvasive placenta accreta spectrum (PAS) diagnosis in pregnant women. Findings MRI features demonstrated variable sensitivity and specificity across readers for PAS diagnosis, often at a trade-off, highlighting the need for consensus review. Clinical relevance A model based on consensus interpretation of MRI findings, comprised of "uterine/placental bulge," "exophytic placental mass," and "bladder vessel sign," demonstrated good performance for diagnosing PAS, which is essential given the risk of maternal morbidity and mortality at delivery.