Accuracy of Post-Neoadjuvant Therapy MRI for the Assessment of Anal Sphincter Involvement in Patients with Rectal...

Purpose To assess the accuracy of post-neoadjuvant therapy (NAT) MRI, as compared with that of pathologic evaluation, to determine anal sphincter involvement in patients with rectal cancer. Materials and Methods This retrospective study included patients diagnosed with rectal cancer between January 2015 and December 2017 whose baseline MRI showed anal sphincter involvement and who then underwent NAT, post-NAT MRI, and abdominoperineal resection. Four radiologists (with 20 years, 5 years, 2 years, and 1 year of experience) independently reviewed MRI findings. Resected specimens were reviewed by a gastrointestinal pathologist. Interreader agreement between the radiologists and pathologist was assessed using the Cohen κ statistic. Conditional sensitivity, specificity, and positive predictive value (PPV) of the radiologists were calculated among patients for whom the radiologists and the pathologist agreed that the anal canal was involved. Results Thirty-two patients were included (mean age ± SD, 60 years ± 15; 19 male, 13 female). For the post-NAT assessment of anal sphincter involvement, agreement between readers 1, 2, and 4 and the pathologist was moderate (κ = 0.55 [95% CI: 0.18, 0.91], 0.45 [95% CI: -0.06, 0.82], and 0.53 [95% CI: 0, 0.89], respectively). There was fair agreement between reader 3 and the pathologist (κ = 0.30 [95% CI: -0.09, 0.67]). Radiologists had high sensitivity for the detection of anal sphincter involvement (88%-100%), high PPV (88%-96%), and moderate to high specificity (50%-80%); the senior radiologist had the highest sensitivity, PPV, and specificity. Conclusion Radiologists had fair to moderate interreader agreement with the pathologist for post-NAT assessment of anal sphincter involvement in patients with rectal cancer and showed high conditional sensitivity regardless of their level of experience. Keywords: Abdomen/GI, Rectum, Oncology, Post-Neoadjuvant Therapy MRI Supplemental material is available for this article. © RSNA, 2025.