Accuracy and Interobserver Reliability of Magnetic Resonance Imaging for Placenta Accreta Spectrum Disorders

Abstract

OBJECTIVE: This study aims to define the accuracy, predictive value, and interobserver reliability of magnetic resonance imaging (MRI) in the diagnosis of placenta accreta spectrum (PAS) disorders. STUDY DESIGN: Two experienced radiologists independently interpreted the MRI studies of patients with possible PAS from two referral centers. Radiologists were blinded to sonographic and clinical information. We calculated diagnostic testing characteristics and kappa statistics of interobserver reliability for MRI findings of PAS. RESULTS: Sixty-eight MRI cases were evaluated. Confirmed PAS and severe PAS were present in 44 (65%) and 20 (29%) cases. For the diagnosis of any PAS, MRI had a sensitivity 66%, specificity 71%, positive predictive value (PPV) 81%, negative predictive value (NPV) 53%, and accuracy 68%. For the diagnosis of severe PAS (percreta), MRI had a sensitivity 85%, specificity 79%, PPV 63%, NPV 93%, and accuracy 81%. The accuracy of individual signs of PAS was lower (44-65%). Interobserver agreement was almost perfect for previa; substantial for myometrial interruptions, PAS, severe PAS, and placental bulging/balling; and moderate to slight for other signs of PAS. CONCLUSION: Although the interobserver reliability of MRI for a diagnosis of PAS is substantial, the accuracy and predictive value are modest and lower than previously reported.