Management of muscle invasive bladder cancer (MIBC) involves neoadjuvant cisplatin-based chemotherapy (NAC) followed by radical cystectomy (RC) for patients who are eligible for both treatments [1,2]. However, NAC offers an absolute overall survival (OS) benefit of only 5% [3], [4], [5] and global utilization remains low [6,7]. Patient risk stratification to determine the indication for NAC is currently based on limited histopathological examination of the tissue from transurethral resection of the bladder tumor (TURBT) and radiological staging [8].

Molecular subtyping has emerged as a potential tool to refine patient selection for NAC. Tumors of certain subtypes have been observed to have a differential response to NAC [9], [10], [11], [12], [13], [14]. For example, Seiler et al. developed a Genomic Subtyping Classifier (GSC) with 4 unique subtypes where the GSC-Basal tumors appeared to benefit the most from NAC, while patients with GSC-Luminal tumors did not [11,15]. In a recent study by Sjödahl et al., tumors assigned to the Genomically Unstable (GU) and Urothelial-like (UroC) Lund subtypes had higher proportions of pathological complete response (pCR) and were associated with better cancer specific survival (CSS) [12]. There are, however, significant inconsistencies between reports linking patient outcomes to molecular subtype.

One likely reason for these inconsistencies is the difficulty in alignment of subtypes between individual models. In part, the discrepancies between subtyping models may be due to differences in nomenclature, where 2 different classifiers may use similar names for a given subtype but not identify the same tumors [16]. In a critical step toward providing a universal reference for molecular subtyping, the Consensus Classifier was developed [14]. However, the clinical utility of this reference classifier remains to be elucidated and supported with prospective trials [17,18].

In a previous study by Lotan et al., we used the Decipher GSC classifier to evaluate the impact of molecular subtypes on OS when controlling for NAC in patients with MIBC and found that patients with GSC-Luminal tumors did not benefit from NAC while the 3 nonluminal subtypes (GSC-Basal, GSC-Claudin Low and GSC-Infiltrated Luminal) received the greatest benefit (10%) from NAC [13]. In this secondary analysis, we have aligned the best characterized molecular subtyping models including the Consensus Classifier, The Cancer Genome Atlas (TCGA) Classifier, and the Lund Classifier, and evaluated their performance in the same meta-cohort. In addition, we explored selected gene expression signatures to delineate the differences between each classifier.