Abstract
INTRODUCTION: Magnetic Resonance Imaging (MRI) is the standard pretreatment staging in patients with rectal cancer. Accurate tumor staging is paramount to determining the appropriate treatment course for patients diagnosed with rectal cancer. The current study aims to re-evaluate the accuracy of pre-operative MRI in staging of both early and locally advanced rectal cancer following completion of neoadjuvant therapy (NAT) compared to the pathologic stage.
METHODS: A retrospective review of patients treated for rectal cancer between 2015 and 2020 at a single academic institution. All patients underwent rectal cancer protocol MRIs before surgical resection. Analysis was carried out in two groups: early rectal cancer: T1/2 N0 tumors with upfront surgical resection (N = 40); and locally advanced disease: T3 or greater or N+ disease receiving NAT, with restaging MRI following NAT (n = 63).
RESULTS: 103 patients were included in analysis. MRI accuracy in early tumors was 35% ICC = 0.52 (95% CI 0.25-0.71) T stage and 66% ICC = 0 (95% CI -0.24, 0.29) for 29 patients with nodal data for N stage. There was 28% understaging of T2 tumors and 34% understaging of N0 stage by MRI. Post NAT MRI had 44% accuracy ICC = 0.57 (95% CI -0.15-0.20) T stage and 60% accuracy ICC = 0.32 (95% CI 0.08-0.52) N stage. Tumor invasion was overstaged on MRI: 40% T2, 29% T3, 90% T4. Nodal inaccuracy was due to overstaging, 61% N1, 90% N2.
CONCLUSIONS: In locally advanced rectal cancer MRI overstaged tumors, this could be due to the continued effect of NAT from MRI to resection. This overstaging is of little clinical significance as it doesn't alter the treatment plan, except in cases of complete clinical response. In early rectal cancer, MRI had limited accuracy compared to pathology, understaging a quarter of patients who would benefit from NAT before surgery. Other adjunct imaging modalities should be considered to improve accuracy in staging early rectal cancer and consideration of complete response and enrollment in watch and wait protocols.