EURURO Vol. 72 No. 4

dataset that combined patients from TCGA, MDA, and the

University of Lund datasets (

= 476). We validated the

relative changes in outcome by GSC subtype in patients

treated with and without NAC

( Fig. 4

A). Patients with GSC

basal tumors had a 3-yr OS rate of 49.2% (95% confidence

interval [CI] 39.5–61.2%;

0.001) in the non-NAC cohort

compared with 77.8% (95% CI 67.2–90.0%;

0.001) in the

NAC cohort.

had the worst outcome, the prognosis of patients with basal tumors significantly improved when treated with NAC. (B) OS stratified according to the

MDA subtypes. Patients with MDA luminal tumors had the best outcome in both the non-NAC (left) and NAC (right) settings. In the presence of NAC,

patients with p53-like tumors had a significantly shorter OS when compared with patients with MDA luminal tumors. (C) OS stratified according to

the TCGA clusters. Clusters I and II clearly subdivide luminal tumors into two subsets—a subset with good prognosis (cluster I) and a subset with poor

prognosis (cluster II)—although neither was affected by NAC. Basal tumors were subdivided into two subsets with similar prognosis in the non-NAC

setting, but discrepant responses to NAC. The OS of patients with cluster III tumors was superior when treated with NAC, whereas that of patients with

cluster IV tumors was poor regardless of NAC. (D) OS stratified according to the Lund subtypes. Patients with luminal tumors (Uro A and genomically

unstable) had the best outcome without (left) and with (right) NAC. The OS of patients with basal tumors (Uro B and SCC-like) was inferior to that of

Uro A tumors in the absence of NAC (left). However, with NAC (right) the outcome was similar to that in Uro A patients. The

values represent Cox

proportional hazard ratios. MDA = MD Anderson Cancer Center; NAC = neoadjuvant chemotherapy; OS = overall survival; SCC = squamous cell

carcinoma; TCGA = The Cancer Genome Atlas; UNC = University of North Carolina; Uro = urobasal.

[(Fig._3)TD$FIG]

Fig. 3 – Discovery and validation of the GSC. (A) Proposed GSC bladder cancer classes derived from a consensus of four models (UNC, MDA, TCGA, and

Lund). Colors indicate each class: claudin-low (gray), basal (red), luminal-infiltrated (light blue), and luminal (dark blue). (B) GSC cross-validation

scores for each sample in the NAC dataset. The vertical bands represent the probability of each sample belonging to each class. The bottom bars

indicate the classes predicted by GSC as well as the consensus classes. (C) Performance of the GSC in the discovery (10-fold cross validation for model

performance) and two independent validation cohorts (NAC validation cohort and non-NAC validation cohort). Across all the cohorts, GSC was able to

predict all subtypes significantly with a high area under the curve (compared with consensus classes). AUC = area under the curve; CI = confidence

interval; Cons. = consensus; GSC = genomic subtyping classifier; inf = infiltrated; MDA = MD Anderson Cancer Center; NAC = neoadjuvant chemotherapy;

TCGA = The Cancer Genome Atlas; UNC = University of North Carolina.

E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 5 4 4 – 5 5 4

550