Endoscopic Feature and Response Reproducibility in Tumor Assessment after Neoadjuvant Therapy for Rectal...

The watch-and-wait approach may be safe for selected rectal cancer patients who achieve a complete clinical response after neoadjuvant treatment. Endoscopic examination is critical in determining completeness of tumor response but has not been systematically studied.

Two cross-sectional surveys, each containing endoscopic photos of rectal cancers treated with neoadjuvant therapy, were distributed to surgeons. The first survey assessed the reproducibility of eight endoscopic criteria using 41 unique endoscopic photos. The percentage of surgeons selecting each of the prespecified endoscopic criteria for each photo was calculated to determine the reproducibility of endoscopic criteria in assessing treatment and tumor response grade across multiple surgeons. The second survey included endoscopic pairs of pre- and post-neoadjuvant treatment photos of 17 patients. The surgeons were assigned a tumor response grade (clinical complete response [cCR], near complete clinical response [nCR], incomplete [iCR] clinical response), and percentages of correct diagnostic assignment were calculated.

The findings showed significant inter- and intra-surgeon variation in the selection of predefined endoscopic features used to grade tumor response as well as significant inter- and intra-surgeon variation in the selection of the tumor response grade (cCR, nCR, or iCR). However, individual endoscopic features and tumor response grades clustered together, suggesting consistency in tumor response interpretation. Surgeons were more accurate in identifying patients with a complete response (82%) than in identifying patients with an incomplete response (68%).

Despite inter- and intra-surgeon variation, endoscopic features were well-selected in terms of tumor response grade, suggesting consistency in endoscopic interpretation. Surgeons tended to underestimate the degree of tumor response, identifying complete responses more accurately than incomplete responses.