Abstract
OBJECTIVES: To evaluate the prognostic significance of tumour deposit (TD) across different N stages in patients with stage III colon cancer and to develop and validate a novel N staging system that incorporates TD count to improve prediction of cancer-specific survival (CSS). DESIGN: Retrospective cohort study based on population-based data and external validation. SETTING: Secondary and tertiary care settings; data from the SEER database, a population-based cancer registry capturing cancer incidence and survival information (USA, 2010-2017); and a single-centre validation cohort from South China (2015-2019). PARTICIPANTS: A total of 8739 stage III colon cancer patients from the SEER database who underwent curative surgery were included; 1335 (15.3%) had TD. Patients with <12 examined lymph nodes and those who received neoadjuvant therapy were excluded. An independent external validation cohort included 326 patients from Guangdong Provincial People's Hospital. Selection criteria were consistent with the SEER cohort. OUTCOME MEASURES: The outcome was cancer-specific survival (CSS). The prognostic impact of tumour deposit (TD) and the comparative performance of the novel N staging system versus the AJCC system were evaluated using the Fine-Gray competing risks model, time-dependent area under the curve (AUC) and Brier score. RESULTS: TD was independently associated with poorer CSS: 1 TD (SHR=1.23, 95% CI 1.04 to 1.47, p=0.017), 2-3 TDs (SHR=1.36, 95% CI 1.17 to 1.58, p<0.001) and >3 TDs (SHR=2.02, 95% CI 1.73 to 2.35, p<0.001). The adverse effect of TD was more pronounced in N1 compared with N2 patients. Based on these findings and the work of Mayo et al, TDs were converted to metastatic lymph nodes (mLNs) using the following weighting: 1 TD=3 mLNs (N1c), 1 TD=2 mLNs (N1) and 1 TD=1 mLN (N2). The novel N staging system stratified patients as nN1a (1 nLN), nN1b (2-3 nLNs), nN2a (4-6 nLNs) and nN2b (≥7 nLNs). This approach showed improved prognostic accuracy compared with AJCC N staging: 3-year AUC (0.623 vs 0.614) and Brier score (0.151 vs 0.157, p<0.001). A similar performance was observed in the external validation cohort: 3-year AUC=0.709 and 5-year AUC=0.684. CONCLUSIONS: TD significantly worsens prognosis in stage III colon cancer, particularly in lower N stages. Incorporating TD counts into the N staging system with different weightings based on N stage enhances prognostic accuracy and risk stratification within stage III disease, particularly for the heterogeneous AJC N1c category. This novel staging system provides better prognostic value and more accurate treatment guidance and should be considered for broader clinical use, subject to further (eg, prospective) validation.