Accuracy of MRI in Restaging Locally Advanced Rectal Cancer After Preoperative Chemoradiation

BACKGROUND: Patients with a locally advanced rectal carcinoma benefit from preoperative chemoradiotherapy. MRI is considered the first choice imaging modality after preoperative chemoradiation, although its reliability for restaging is debatable. OBJECTIVE: The purpose of this study was to determine the accuracy of MRI in restaging locally advanced rectal cancer after preoperative chemoradiation. DESIGN: This was a retrospective study. SETTINGS: The study was conducted in a Dutch high-volume rectal cancer center. PATIENTS: A consecutive cohort of 48 patients with locally advanced rectal cancer treated with a curative intent was identified. MAIN OUTCOME MEASURES: Three readers independently evaluated the MRI both for primary staging and for restaging after preoperative chemoradiation and were blinded to results from the other readers as well as histological results. Interobserver variability was determined. Accuracy of the restaging MRI was assessed through the comparison of tumor characteristics on MRI with histopathologic outcomes. RESULTS: T stage was correctly predicted by the 3 readers in 47% to 68% and N stage in 68% to 70%. Overstaging was more common than understaging. Positive predictive values (PPV) among the 3 readers for T0 were 0%, and negative predictive values (NPVs) varied from 84% to 85%. For T1/2, PPVs and NPVs were 50% to 67% and 72% to 90%, and for T3/4 they were 54% to 62% and 33% to 78%. PPVs and NPVs for N0 stage were 81% to 95% and 58% to 73%. Tumor regression grade on MRI did not correspond with histopathologic tumor regression grade; PPVs for good response (tumor regression grade on MRI 1-2) were 48% to 61%, and NPVs were 42% to 58%. Interobserver agreement was fair to moderate for T stage, N stage, and tumor response (κ = 0.20-0.41) and fair to substantial for the relation with the mesorectal fascia (κ = 0.33-0.77). In none of the patients was the surgical plan changed after the restaging MRI. LIMITATIONS: This study was limited by its small sample size and retrospective nature. CONCLUSIONS: MRI has low accuracy for restaging locally advanced rectal cancer after preoperative chemoradiation, and the interobserver variability is significant.