- Design
- Single-centre retrospective diagnostic accuracy study with head-to-head comparison and DeLong testing
- Population
- 136 patients with 138 pathologically confirmed Tis-T2 rectal lesions (82 Tis-T1, 56 T2) resected 2020-2023; mean age 60, 78 men
- Primary outcome
- Accuracy of MRI features and endorectal ultrasound for identifying stage Tis-T1 disease
- Effect
- AUC 0.762 muscularis propria status, 0.861 submucosal enhancing stripe, 0.915 combined model, 0.806 endorectal ultrasound; combined vs ultrasound difference 0.109 (p=0.003)
Whether an early rectal cancer is Tis-T1 or T2 decides whether local excision is an option, and the two are hard to separate. Endorectal ultrasound is the traditional answer and is operator-dependent. This study asked whether adding contrast-enhanced MRI sequences does better.
136 patients with 138 pathologically confirmed Tis-T2 lesions (82 Tis-T1, 56 T2), all resected between 2020 and 2023, had preoperative MRI including contrast-enhanced sequences and endorectal ultrasound. A radiologist assessed tumour shape and the status of the muscularis propria on T2-weighted images, and the submucosal enhancing stripe on the contrast-enhanced sequences; an endoscopist staged by ultrasound.
Areas under the curve for identifying Tis-T1 were 0.762 for muscularis propria status alone, 0.861 for the submucosal enhancing stripe, 0.915 for the combined model, and 0.806 for endorectal ultrasound. The combined model beat the T2-only approach by 0.154 (p<0.001) and beat ultrasound by 0.109 (p=0.003).
The caveats are the usual ones for a single-centre diagnostic study: one radiologist assessed the MRI features, so inter-reader reproducibility of the submucosal enhancing stripe is unknown and is exactly what would determine whether this transfers; and all patients went to surgery, which selects the population.
Where it points is worth acting on cautiously. If a department already acquires contrast-enhanced sequences in rectal MRI, the submucosal enhancing stripe is a feature worth learning and reporting. It is not yet a reason to add contrast where the protocol does not include it, and organ-sparing decisions should still be made in a multidisciplinary meeting with the endoscopic appearance in front of it.
- Report the submucosal enhancing stripe where contrast-enhanced sequences are already acquired
- Do not add contrast to a rectal MRI protocol on a single-centre study with one reader
- Keep the endoscopic appearance in the multidisciplinary discussion - imaging alone does not decide local excision
- Note the selection: every patient here went to curative resection
- Inter-reader reproducibility of the new feature has not been established
The statistics, in plain English
An area under the curve of 0.915 against 0.806 means the combined MRI model ranks a Tis-T1 lesion above a T2 lesion about 92% of the time versus 81% for ultrasound - a real difference, and the DeLong test confirms it is unlikely to be chance. With 138 lesions, however, the intervals are wide enough (0.856-0.956) that the true advantage could be smaller than it appears. A single reader assessing the imaging features gives the best-case estimate of performance: in routine practice, with readers of varying experience, accuracy is almost always lower.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for radiology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free