Combination of a Novel Gene Expression Signature with a Clinical Nomogram Improves the Prediction of Survival in High-Risk Bladder Cancer

Purpose: We aimed to validate and improve prognostic signatures for high-risk urothelial carcinoma of the bladder. Experimental Design: We evaluated microarray data from 93 patients with bladder cancer managed by radical cystectomy to determine gene expression patterns associated with clinical and prognostic variables. We compared our results with published bladder cancer microarray data sets comprising 578 additional patients and with 49 published gene signatures from multiple cancer types. Hierarchical clustering was utilized to identify subtypes associated with differences in survival. We then investigated whether the addition of survival-associated gene expression information to a validated postcystectomy nomogram utilizing clinical and pathologic variables improves prediction of recurrence. Results: Multiple markers for muscle invasive disease with highly significant expression differences in multiple data sets were identified, such as fibronectin 1 (FN1), NNMT, POSTN, and SMAD6. We identified signatures associated with pathologic stage and the likelihood of developing metastasis and death from bladder cancer, as well as with two distinct clustering subtypes of bladder cancer. Our novel signature correlated with overall survival in multiple independent data sets, significantly improving the prediction concordance of standard staging in all data sets [mean Delta C-statistic: 0.14; 95% confidence interval (CI), 0.01-0.27; P < 0.001]. Tested in our patient cohort, it significantly enhanced the performance of a postoperative survival nomogram (Delta C-statistic: 0.08, 95% CI, -0.04-0.20; P < 0.005). Conclusions: Prognostic information obtained from gene expression data can aid in posttreatment prediction of bladder cancer recurrence. Our findings require further validation in external cohorts and prospectively in a clinical trial setting. Clin Cancer Res; 18(5); 1323-33. (C)2012 AACR.