CT-based Node-RADS for metastatic lymph node detection in colon cancer and influence of microsatellite instability.
[PURPOSE] To evaluate the diagnostic accuracy of the CT-based Node-RADS score in differentiating benign and metastatic lymph nodes in colon cancer compared with standard size parameters and to assess
- p-value p < 0.001
- 95% CI 2.46-12.05
APA
Conca M, Rodà GM, et al. (2026). CT-based Node-RADS for metastatic lymph node detection in colon cancer and influence of microsatellite instability.. Abdominal radiology (New York). https://doi.org/10.1007/s00261-026-05477-2
MLA
Conca M, et al.. "CT-based Node-RADS for metastatic lymph node detection in colon cancer and influence of microsatellite instability.." Abdominal radiology (New York), 2026.
PMID
41961314
Abstract
[PURPOSE] To evaluate the diagnostic accuracy of the CT-based Node-RADS score in differentiating benign and metastatic lymph nodes in colon cancer compared with standard size parameters and to assess the influence of microsatellite instability on diagnostic performance.
[METHODS] This retrospective study included 166 patients with histologically confirmed colon cancer who underwent preoperative contrast-enhanced CT. Two radiologists, blinded to histopathological findings, evaluated a total of 166 lymph nodes under the supervision of a third radiologist, calculating Node-RADS scores and long and short diameters. Univariate and multivariate analyses were performed, stratifying data based on histopathological findings and microsatellite instability. Diagnostic performance for the various radiological variables was assessed using Receiver Operating Characteristic (ROC) curves. Predictive logistic models were created.
[RESULTS] Sixty-nine lymph nodes (42%) were histologically positive. Interobserver agreement for the Node-RADS score was excellent (κ = 0.88). All Node-RADS features were significantly associated with nodal status (p < 0.001), with texture showing the strongest correlation (Cramér's V = 0.74) and emerging as the only independent predictor in multivariable analysis (OR 5.10, 95% CI 2.46-12.05, p < 0.001). Overall, the Node-RADS score demonstrated good discriminative ability (AUC = 0.79, 95% CI 0.72-0.86), although not superior to short axis diameter (AUC = 0.75, 95% CI 0.68-0.83). Optimal diagnostic accuracy (0.77, 95% CI 0.70-0.93) was observed using a cutoff between Node-RADS categories 3 and 4. In tumors with microsatellite instability, Node-RADS performance decreased (AUC = 0.42, 95% CI 0.19-0.66) compared with microsatellite-stable tumors (AUC = 0.83, 95% CI 0.76-0.9). Repeated 5-fold cross-validation showed that the composite Node-RADS score was not inferior to logistic regression models including individual imaging features (AUC 0.786-0.802; all DeLong p > 0.01).
[CONCLUSION] The Node-RADS score demonstrates good diagnostic performance for nodal staging in colon cancer; however, its superiority over standard size criteria was not confirmed. Its diagnostic performance declines in tumors with microsatellite instability, underscoring the potential influence of microsatellite status on the reliability of CT-based lymph node assessment.
[METHODS] This retrospective study included 166 patients with histologically confirmed colon cancer who underwent preoperative contrast-enhanced CT. Two radiologists, blinded to histopathological findings, evaluated a total of 166 lymph nodes under the supervision of a third radiologist, calculating Node-RADS scores and long and short diameters. Univariate and multivariate analyses were performed, stratifying data based on histopathological findings and microsatellite instability. Diagnostic performance for the various radiological variables was assessed using Receiver Operating Characteristic (ROC) curves. Predictive logistic models were created.
[RESULTS] Sixty-nine lymph nodes (42%) were histologically positive. Interobserver agreement for the Node-RADS score was excellent (κ = 0.88). All Node-RADS features were significantly associated with nodal status (p < 0.001), with texture showing the strongest correlation (Cramér's V = 0.74) and emerging as the only independent predictor in multivariable analysis (OR 5.10, 95% CI 2.46-12.05, p < 0.001). Overall, the Node-RADS score demonstrated good discriminative ability (AUC = 0.79, 95% CI 0.72-0.86), although not superior to short axis diameter (AUC = 0.75, 95% CI 0.68-0.83). Optimal diagnostic accuracy (0.77, 95% CI 0.70-0.93) was observed using a cutoff between Node-RADS categories 3 and 4. In tumors with microsatellite instability, Node-RADS performance decreased (AUC = 0.42, 95% CI 0.19-0.66) compared with microsatellite-stable tumors (AUC = 0.83, 95% CI 0.76-0.9). Repeated 5-fold cross-validation showed that the composite Node-RADS score was not inferior to logistic regression models including individual imaging features (AUC 0.786-0.802; all DeLong p > 0.01).
[CONCLUSION] The Node-RADS score demonstrates good diagnostic performance for nodal staging in colon cancer; however, its superiority over standard size criteria was not confirmed. Its diagnostic performance declines in tumors with microsatellite instability, underscoring the potential influence of microsatellite status on the reliability of CT-based lymph node assessment.