The Vanderbilt staging system for retroperitoneal sarcoma: a validation study of 6857 patients from the National Cancer Database

Abstract

The Vanderbilt staging system for retroperitoneal sarcoma incorporates information regarding the histologic subtype of sarcoma and outperforms the 7th and 8th editions of the AJCC staging systems by several different statistical criteria. The current study was performed using an independent patient cohort from the National Cancer Database (n = 6857) to validate this proposed staging system. Each staging system was assessed for degree of discrimination by pairwise comparisons of adjacent stage categories. Predictive accuracy of 5-year overall survival was performed by comparison of areas under receiver operating characteristic curves generated from logistic regression. Three different concordance indices (Harrell's c, Somers' D, and Gönen and Heller's K) were calculated using bootstrap methods. Amount of variation in observed outcomes explained by each staging system was assessed using O'Quigley's ρ2k and Royston's R2. Bayesian information criteria were also assessed as measures of model fit. The revised AJCC 8th edition T categories were not effective in categorizing risk of death. The Vanderbilt staging system showed the best discrimination between adjacent tumor stages, highest predictive accuracy for 5-year overall survival, a higher degree of concordance with and explained variation of clinical outcomes, and resulted in the best fitting regression model. These results obtained with an independent dataset validate the Vanderbilt staging system for retroperitoneal sarcoma and demonstrate its superiority in risk stratification over current and prior editions of the AJCC staging system.