Nonmuscle invasive bladder cancer (NMIBC) accounts for the majority of newly diagnosed bladder cancer cases in the U.S. These tumors exhibit a heterogeneous range of outcomes along the spectrum of cancer recurrence and progression and this underscores the complexity of clinical management. To standardize treatment, various risk stratification systems were developed to guide clinicians. Risk stratification tools include the European Organization for Research and Treatment of Cancer (EORTC),¹ the Spanish Urological Club for Oncological Treatment (CUETO),² the European Association of Urology (EAU)³ and American Urological Association/Society of Urologic Oncology (AUA/SUO) models.⁴ The EORTC and CUETO models are scoring systems that use clinical parameters to calculate the probability of recurrence and progression and are based on clinical trial data. The EAU and AUA/SUO risk stratification tools are broadly derived from the EORTC and CUETO models but are simplified into low, intermediate and high-risk categories (see Appendix).

Current Risk Stratification

Although the EORTC and CUETO models are based on data from clinical trials, there were notable differences in the management of these patients compared to current guideline recommendations. Specifically, EORTC trial patients were mainly treated with intravesical chemotherapy, and tumor grade was classified using the 1973 World Health Organization (WHO) classification whereas CUETO trial patients received mainly bacillus Calmette-Guérin (BCG). Additionally, patients used in both models did not routinely undergo repeat transurethral resection for high-grade tumors, immediate instillation of post operative chemotherapy, or receive long-term maintenance therapy. Therefore, the predicted risk of recurrence and progression may be overestimated. Recently, the EORTC model was updated to include BCG-treated patients and the EAU updated their progression risk model to include the WHO 2004/2016 grading classification.⁵

The EAU and AUA guidelines risk groupings are based on expert consensus of clinical parameters to categorize patients. However, there are several important differences between them. For example, the AUA considers solitary, small (<3 cm), high-grade tumors as intermediate-risk whereas EAU considers these high-grade tumors as high-risk. The EAU also added a very high-risk group to emphasize the adverse outcome for those with multiple high-risk features.⁵ The AUA risk stratification includes other clinical variables like lymphovascular invasion, prostatic urethral involvement, variant histology, and poor response to BCG and prior recurrence to better categorize patients, and our group has shown that the AUA risk group appropriately stratifies patients when applied to a contemporary NMIBC cohort.⁶

Clinical Web-Based Applications

Of the utmost importance to clinicians is the ease of use and applicability of these risk stratification tools and therefore they need to be readily available in the clinical setting. The EAU (www.nmibc.net), EORTC (www.eortc.be/tools/bladdercalculator) and CUETO (www.aeu.es/Cueto.html) have each created free online web applications for risk calculation.

Web applications may also incorporate risk tools into a comprehensive guideline application that can walk the user through the guideline algorithm in a concise and user-friendly manner. One such example is the online web application BLATUR (www.blatur.com), which was derived from the AUA guidelines, and utilizes clinical data input by the user to risk stratify patients and then recommend the next step in management (fig. 1). Given the predominance of smartphones and electronic medical records, integration of these applications and calculators into standard practice and related documentation offers the potential benefit of improving adherence to guidelines and may lead to better outcomes.

Figure 1. BLATUR online web application (www.blatur.com).

Role of Imaging

Ongoing work is underway with multiparametric magnetic resonance imaging (mpMRI) to assess the bladder and provide clinical staging of NMIBC using the Vesical Imaging Reporting and Data System (VI-RADS) score (fig. 2). Recently, Del Diudice et al demonstrated that the VI-RADS score can differentiate NMIBC from muscle invasive bladder cancer (MIBC) preoperatively and may identify high-risk NMIBC patients who might benefit from, or potentially can avoid, re-transurethral resection of bladder tumor (re-TUR).⁷ By studying 231 patients, the authors revealed that the VI-RADS AUC was 0.94 to discriminate between NMIBC and MIBC preoperatively. In addition, they showed that mpMRI had a 74.5% positive predictive value and 96.6% negative predictive value for identifying MIBC at re-TUR.⁷ Thus the VI-RADS score may be useful in identifying invasive disease and could therefore enhance our ability to risk stratify and distinguish very high-risk NMIBC from MIBC patients.

Figure 2. Preoperative mpMRI of female patient with bladder cancer with tumor extension into muscularis propria in T2 weighted imaging (A), diffusion weighted imaging (B), and apparent diffusion coefficient maps with significantly restricted diffusion (C). Dynamic contrast enhanced imaging (D) and perfusion map (E) revealed early and heterogeneous lesion enhancement. VI-RADS score given for this patient is 4. Figure adapted from Del Giudice et al with permission.⁷

Molecular and Genomic Profiling

Figure 3. Summary characteristics of transcriptomic classes. Figure adapted from Lindskrog et al with permission.⁹

Currently, there are no guideline-based risk tools that utilize molecular or genomic profiles to risk stratify patients. However, ongoing research is underway to identify pathways and markers that may predict treatment outcomes. In a study of high-risk NMIBC patients, Meeks et al demonstrated that tumor mutational burden was different between progressors and nonprogressors, and also identified a possible association between increased frequency of loss of CDKN2A/B and progressors versus nonprogressors.⁸ Lindskrog et al published a comprehensive analysis of multi-omics in NMIBC and associated outcomes of 834 patients enrolled in the UROMOL project.⁹ They discovered that NMIBC aggressiveness is associated with genomic alterations, immune cell infiltration, and transcriptomic classes. Each transcriptomic class was associated with certain gene expression profiles, mutations, and immune activation. Additionally, the authors proposed potential NMIBC therapies for each class based on disease aggressiveness. For example, class 1 NMIBC was associated with FGFR3, and RAS genomic alterations, and relatively lower recurrence rates whereas class 2a NMIBC had high chromosomal instability and poor outcome (fig. 3). Thus, class 1 tumors may be managed with close surveillance and/or intravesical chemotherapy, whereas class 2a tumors may require BCG therapy and even possibly early cystectomy. Moreover, the authors demonstrated that incorporating transcriptomic subtypes into EORTC risk classification improved the AUC from 0.77 to 0.85 (95% CI 0.78–0.91). Additional work on genomic profiling in NMIBC by Pietzak et al showed that NMIBC tumors with ARID1A mutations were at higher risk of recurrence after BCG immunotherapy (HR 3.14, 95% CI 1.51–6.51, p=0.002).¹⁰ Thus, emerging data support the potential role of molecular and genomic profiling in improving our ability to predict the risk of recurrence and progression.

In summary, risk stratification in NMIBC is a crucial first step in any treatment algorithm. While current models are based on sound clinical data, their accuracy has limitations. The incorporation of web-based applications to improve clinician access through risk stratification and adherence to guidelines via smartphone, tablet, and electronic medical records will be valuable for improving the quality of care and treatment outcomes. Enhanced imaging techniques such as mpMRI will add another valuable dimension to risk assessment. Finally, with evolving techniques such as next-generation sequencing and molecular profiling, the integration of NMIBC tumor biology into risk stratification tools may improve predictive ability beyond that of current models.