Introduction

Endoscopic intermuscular dissection (EID) has been shown to be an effective technique to remove deep submucosal invasive rectal cancers. However, selecting tumours suitable for this technique has been challenging both endoscopically and radiologically, due to a lack of knowledge and confidence in the anatomical characterisation of early rectal cancers (ERC). This has limited attempts at rectal-preserving local excision, but has also led to incomplete resection of deeper invasive cancers. We aimed to determine if the application of a novel systematic approach for staging rectal lesions on MRI can more accurately identify tumours suitable for local excision in the intermuscular plane, by identifying preservation of at least 1 mm of muscularis propria on MRI.

Aims & Methods

Twelve Dutch abdominal radiologists participated in a one-day training session by an expert radiologist who developed and validated the systematic reporting approach for assessing ERC on MRI. After training, 12 study radiologists and 1 expert radiologist reported a retrospective cases series of pseudonymised MRI scans, blinded to pathological outcome data. Scans were derived from consecutive patients from 2 academic centers in the Netherlands who were suspected of having (at least) deep submucosal invasive rectal cancer at optical diagnosis between 2018 and 2022. All scans were matched to clinical and pathological outcome data. Our primary outcome was the accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of MRI in identifying tumours suitable for a local excision in the intermuscular plane, based on preservation of ≥ 1 mm of muscularis propria on MRI, and compared to pathological staging (≤ pT2 limited to the circular m. propria). Secondary outcome was the inter-observer reliability between study and expert radiologists.

Results

246 patients were included (median age 67 years [IQR 13], 71.1% male, 86.6% ASA I/II), of which 186 (75.6%) would retrospectively have been suitable candidates for a local excision in the intermuscular plane based on the pathology data (≤ pT2 limited to the circular m. propria). MRI accuracy in identifying these tumours as locally resectable was 81.7% (95% CI 76.3 - 86.3) for expert radiologist diagnosis and 78.5% (95% CI 72.7 - 83.5) for the trained study radiologists (range individual radiologists: 64.1% - 94.6%). Sensitivity, specificity, PPV, and NPV were 86.0%, 68.3%, 89.4%, and 61.2% for expert diagnosis, and 80.5%, 72.4%, 90%, and 54.6% for trained study radiologists, respectively. The percentage of locally resectable cancers overstaged by MRI was 14.0% based on expert radiologist assessment and 19.5% for the trained study radiologists, and the percentage of understaged cancers was 31.7% and 27.3%, respectively. Expert and study radiologists showed overall moderate inter-observer reliability (kappa 0.46; 95% CI 0.34 - 0.59). Inter-observer reliability among study radiologists was moderate as well (kappa 0.44; 95% CI 0.32 - 0.57).

Conclusion

MRI has not previously been considered useful for identification of ERC suitable for endoscopic local excision. This study shows our approach to select rectal cancers for excision in the intermuscular plane is accurate, easily taught and reproducible. This can improve care by enabling better case selection for rectal-preserving local excisions such as EID. Further prospective research is needed to evaluate the clinical impact of this approach.