ABSTRACT
BACKGROUND: Lymph node ratio (LNR) is increasingly accepted as a useful prognostic indicator in colorectal cancer. However, variations in methodology, statistical stringency and cohort composition has led to inconsistency in respect of the optimally prognostic LNR. OBJECTIVE: The aim was to apply a robust regression-based analysis to generate and appraise LNRs optimally prognostic for colon and rectal cancer, both separately and in combination. METHODS: LNR was established for all patients undergoing either a colonic (n = 379) or rectal (n = 160) cancer resection with curative intent. The optimal LNR associated with disease-free and overall survival were established using a classification and regression tree technique. This process was repeated separately for patients who underwent either colonic or rectal resection and for the combined cohort. Survival associated with differing LNR was estimated using the Kaplan-Meier method and compared using a log-rank test. Relationships between LNR, disease-free survival (DFS) and overall survival (OS) were further characterised using Cox regression analysis. All statistical analyses were conducted in the R programming environment, with statistical significance was taken at a level of p < 0.05. RESULTS: Optimal LNRs differed between each cohort, when either overall or disease-free survival was considered. LNRs generated from combined cohorts also differed from those generated by individual cohorts. In relation to DFS, LNR values were obtained and included 0.18 for the colon cancer cohort and 0.19 for the rectal and combined colorectal cancer cohorts. In relation to OS, multiple LNR values were obtained for colon and combined cohorts; however, an optimal LNR was not evident in the rectal cancer cohort. Survival patterns according to LNR closely resembled those associated with standard nodal staging. CONCLUSION: Application of a data-driven approach based on recursive partitioning generates differing lymph node ratios for colon, rectal and combined colorectal cohorts. In each cohort, LNR was similarly prognostic to standard nodal staging in respect to overall and disease-free survival. Overall survival was associated with a multiplicity of LNR values, whilst disease-free survival was associated with a single LNR only. The paper demonstrates the merits of utilising a data-driven approach to determining lymph node ratios from specific patient cohorts. Utilising such an approach enabled the generation of those LNRs that were most associated with particular survival trends in relation to overall and disease-free survival. These differed markedly for colon cancer, rectal cancer and combined cohorts. In general, the survival patterns associated with LNRs generated were similar to those observed with standard nodal staging.
