Identifying NE variants may have prognostic implications and modify treatment recommendations. In this study, they present a genomic classifier trained to identify NE-like tumors. Using unsupervised clustering of transcriptome-wide expression profiles, they generated a series of clustering solutions, evaluating the resultant model on a testing cohort (n=225). A GLMnet model was used to train the classifier to identify NE-like tumors in a training cohort (n=173). The classifier was applied to 4 validation cohorts (n=1030).
Survival analyses were used to characterize the clinical outcomes of the NE-like tumors. In the training set, hierarchical clustering using a panel of 84 genes showed a cluster of 17 patients (9.8%) with highly heterogeneous expression of NE markers but no expression of basal or luminal markers. This biological profile was consistently seen across the 4 validation cohorts. These patients had significantly worse 1 year progression free survival (65% vs 82% for NE-like vs overall; p=0.046). After adjusting for various clinical and pathological factors, patients with NE-like tumors had 6 times increased risk of all-cause mortality (p=0.001) than non-NE tumors.
In summary, they developed a single patient classifier that identifies a particularly high-risk group of NE-like tumors that may need treatment intensification, alternative chemotherapy or clinical trials. Further validation in larger cohort will be required to assess potential clinical utility.
Presented by: Peter C. Black, University of British Columbia, Vancouver, Canada
Written by: Stephen B. Williams, M.D., Associate Professor, Division of Urology, The University of Texas Medical Branch, Galveston, TX. and Ashish M. Kamat, M.D. Professor, Department of Urology, Division of Surgery, The University of Texas MD Anderson Cancer Center, Houston, TX at the 16th Annual Meeting of the International Bladder Cancer Network (IBCN) October 11-13, 2018 - the Inntel Hotels Rotterdam Centre, Rotterdam, The Netherlands