Prognostic accuracy of different lymph node staging systems in rectal adenocarcinoma with or without preoperative radiation therapy

It is still controversial about the performance of different lymph node stating systems in rectal cancer. LODDS and LNR should be considered when assessing the prognosis of rectal adenocarcinoma patients. Abstract Background and objective A variety of different lymph node (LN) staging systems have been developed to describe the lymph node status accurately. We aim to compare the prognostic accuracy of American Joint Committee on Cancer seventh N stage relative to negative number of lymph node (nLN), lymph node ratio (LNR) and log odds of metastatic lymph nodes (LODDS) in rectal adenocarcinoma (RC). Methods A total of 19 167 Stage II–III rectal cancer patients who underwent surgical resection of rectal adenocarcinoma were identified from Surveillance, Epidemiology and End Results database. Akaike’s Information Criterion (AIC) and the Harrell’s concordance index (c statistic) were used to evaluate the relative discriminative power of the different LN staging systems. Results Of the 19 167 patients, 10 958 received preoperative radiotherapy (pre-RT cohort) and 8209 patients were treated with surgical resection directly (SURG cohort). When assessed using categorical cutoff values, LNR has a somewhat better prognostic accuracy both in pre-RT (c-index: 0.62; AIC: 2988.6) and SURG groups (c-index: 0.60; AIC: 3359.8). Further analysis based on different total number of lymph node (TNLN) suggested that when less than 10 lymph nodes were retrieved, LNR exhibited significant superiority (pre-RT: c-index: 0.597, AIC: 1006.8; SURG: c-index: 0.560, AIC: 810.5). When analyzed as a continuous variable, the LODDS system performed the best and was not impacted by TNLN. Conclusion When assessed as a categorical variable, LNR was the most powerful method to predict survival for Stage II–III RC patients with limited TNLN. Rather, LODDS was the most accurate staging system regardless of the TNLN when LN status was modeled as continuous variable.