A Novel Quantitative Endoscopic Tumor Regression Grading System for Predicting Pathological Response Following...

While pathological and magnetic resonance imaging-based tumor regression grading systems are widely utilized following neoadjuvant therapy for rectal cancer, existing endoscopic approaches rely on qualitative assessment without quantitative weighting of individual features. Although individual endoscopic features associated with response have been described, no weighted scoring system exists to integrate these features into an objective assessment tool. This lack of a statistically weighted grading framework leads to subjective clinical impressions and significant interobserver variability in documentation and interpretation, limiting standardized patient selection for organ preservation strategies.

To develop a quantitative, point-based endoscopic tumor Regression Grading system to standardize post-neoadjuvant therapy endoscopic assessment and predict pathological complete response.

Retrospective cohort study.

Single tertiary care center.

Patients with locally advanced rectal cancer treated with total neoadjuvant therapy and surgical resection between 2018 and 2025 who had complete pre- and post-total neoadjuvant therapy endoscopic documentation were included.

Two experienced colorectal surgeons independently evaluated endoscopic images in blinded fashion, scoring predefined endoscopic features. A statistically weighted, point-based scoring algorithm was developed, with quantitative point values assigned proportional to the strength of association with American Joint Committee on Cancer pathological response. The system incorporated mucosal healing features (whitening, telangiectasia, scarring, erythema), adverse features (ulceration, necrosis), mass regression levels, and stricture characteristics, classifying patients into 3 tiers: Endoscopic Tumor Regression Grading 1 (complete), Endoscopic Tumor Regression Grading 2 (near complete), or Endoscopic Tumor Regression Grading 3 (partial/incomplete).

Predictive performance of the Endoscopic Tumor Regression Grading system for pathological complete response, 3-tier risk stratification, and interobserver reliability, evaluated using diagnostic performance metrics and receiver operating characteristic analysis.

Seventy-six patients (mean age 58 ± 12 years; 58% male) were eligible. The 2 endoscopists demonstrated excellent interobserver agreement for the overall response assessment (weighted κ = 0.9) with 88.2% exact agreement. Univariate analysis identified whitening of the mucosa as the strongest predictor of complete response (odds ratio 47), followed by telangiectasia (odds ratio 23), scarring (odds ratio 16), and erythema (odds ratio 16). Ulceration and necrosis were powerful negative predictors (odds ratio 0.03 each). Each incremental level of mass regression (from no regression to < 50% to 50-99% to 100% regression) increased the odds of complete response 9.6-fold (all p < 0.01). The Endoscopic Tumor Regression Grading system demonstrated strong discrimination for predicting pathological complete response (area under the curve 0.85, 95% confidence interval 0.76-0.93) with a sensitivity of 83.3%, specificity of 76.1%, and accuracy of 78.9%. The Endoscopic Tumor Regression Grading 1 category identified patients with 69.4% pathological complete response probability, while Endoscopic Tumor Regression Grading was associated with 85.7% probability of pathological partial or incomplete response.

Retrospective single-center pilot study requiring external validation.

The Endoscopic Tumor Regression Grading system is the first statistically weighted, point-based scoring algorithm for endoscopic response assessment following neoadjuvant therapy. By quantitatively integrating multiple concurrent endoscopic features weighted by their predictive strength, this system addresses the limitations of qualitative grading, provides objective 3-tier risk stratification, and facilitates standardized patient selection for organ preservation strategies. See Video Abstract.