Abstract
Background
International Metastatic Renal Cell Carcinoma Database Consortium model predicts the outcomes of metastatic renal cell carcinoma stratified into favorable, intermediate, and poor risk groups (FG, IG, and PG, respectively), with approximately 50% of patients being classified as IG. We aimed to generate better risk model based on the sub-classification of IG.
Methods
We analyzed records of 213 consecutive patients receiving molecular targeted therapy. Age, gender, histology, type of initial molecular targeted therapy, serum laboratory data, previous nephrectomy and immunotherapy, and metastatic sites were used for IG sub-stratification. Modified and original models were compared using a concordance correlation coefficient analysis.
Results
Median follow-up was 17.8 months. Serum albumin, serum C-reactive protein, and bone metastases were independent predictors of overall survival (OS) in IG. IG was sub-classified into low-, middle-, and high-risk IG according to the number of predictors. The following modified model was developed: modified FG (FG & low-risk IG), modified IG (middle-risk IG), and modified PG (PG & high-risk IG). Concordance indices for original and modified models were 0.68 and 0.73, respectively (P < 0.001). OS was significantly longer in modified PG treated with mammalian target of rapamycin inhibitors as second-line therapy than with tyrosine kinase inhibitors, whereas this was not observed in the original model.
Conclusions
We successfully developed modified IMDC model using a two-step process: the original IMDC plus an IG sub-stratification, and demonstrated that it predicts outcomes more accurately than original model.
Automated classification of renal cell carcinoma subtypes using bag-of-features
