Postoperative prediction of disease-free survival in rectal cancer after neoadjuvant chemoradiotherapy and radical resection: a nomogram incorporating tumor regression grade and lymph node ratio

Purpose To develop and validate a nomogram incorporating tumor regression grade (TRG) and lymph node ratio (LNR) for predicting disease-free survival (DFS) after neoadjuvant chemoradiotherapy and radical resection for rectal cancer. Methods This retrospective study included 437 patients randomly divided into training (n = 305) and internal holdout validation (n = 132) cohorts, and 88 patients from an independent hospital for external validation. A nomogram estimating 3-year DFS was constructed using a multivariable Cox model. Performance was assessed using the concordance index (C-index), time-dependent area under the receiver operating characteristic curve (AUC), calibration, and decision curve analysis. Results Median follow-up in the original cohort was 35.38 months. In the training and internal holdout cohorts, 106 patients (24.3%) experienced recurrence, distant metastasis, or death, as did 25 patients (28.4%) in the external validation cohort. In the multivariable model, poorer DFS was associated with TRG 2–3 (hazard ratio [HR], 2.45; 95% CI, 1.29–4.65), ypT3–4 (HR, 2.15; 95% CI, 1.12–4.15), ypN-positive disease (HR, 3.39; 95% CI, 1.73–6.63), and higher LNR (HR per 0.1 increase, 1.18; 95% CI, 1.06–1.32). The C-index was 0.830 (95% CI, 0.787–0.876) in training and 0.805 (95% CI, 0.726–0.887) in internal validation; the optimism-corrected C-index in the training cohort was 0.825. Compared with a reference model containing only ypT and ypN categories, the final model showed higher AUCs in the training cohort (0.841 vs. 0.792; AUC difference, 0.049; 95% CI, 0.025–0.088; P = 0.003), internal holdout validation cohort (0.789 vs. 0.735; AUC difference, 0.054; 95% CI, 0.005–0.108; P = 0.036), and external validation cohort (0.861 vs. 0.819; AUC difference, 0.042; 95% CI, 0.003–0.081; P = 0.034). After Holm adjustment, the P values for the AUC comparisons were 0.009, 0.068, and 0.068 in the training, internal holdout validation, and external validation cohorts, respectively. External calibration showed deviations in some probability ranges; decision curve analysis suggested potential net benefit. Conclusions The four-predictor postoperative nomogram showed promising discrimination for 3-year DFS, while its calibration, clinical utility, and generalizability require further evaluation in larger independent studies.