Risk Stratification in Non-Muscle-Invasive Bladder Cancer in the Era of Expanding Therapeutic Options: Are Current Classifications Still Adequate?

BACKGROUND

NMIBC risk classifications from AFU, EAU, AUA/SUO and NCCN were derived decades ago to estimate recurrence and progression, but the therapeutic landscape has since expanded, gene therapy, an IL-15 superagonist, systemic immunotherapy, and BCG-combination regimens, raising the question of whether these classifications remain adequate to guide treatment selection.

OBJECTIVE

To determine whether contemporary NMIBC risk classifications remain adequate for treatment selection, or function as prognostic instruments applied beyond their original purpose.

DATA SOURCES: PubMed/MEDLINE, AFU/EAU/AUA-SUO/NCCN guideline portals, FDA/EMA databases, and ClinicalTrials.gov, searched from January 2000 to August 2026.

STUDY SELECTION: Narrative review prioritizing original derivation and validation studies of risk models, pivotal randomized trials, and current guideline texts over secondary or conference-only sources, flagged as such throughout.

RESULTS

AFU, EAU, AUA/SUO and NCCN diverge architecturally, producing discordant classifications for identical patients, most sharply for low-burden Ta high-grade disease and T1 disease with concomitant CIS. Four agents now hold FDA approval for BCG-unresponsive disease, and two of three recent phase III trials in BCG-naive high-risk disease (CREST, POTOMAC) were positive, but eligibility depends on treatment-exposure definitions outside baseline risk labels; real-world adherence to adequate BCG is below 50%. MRI/VI-RADS shows strong accuracy for muscle invasion but remains unvalidated for risk stratification. We propose a seven-axis, treatment-oriented framework, reassessed at clinical checkpoints, illustrated with two worked examples.

LIMITATIONS: Narrative, not systematic, review; some evidence is conference-sourced pending peer-reviewed publication, and the framework is not yet validated.

CONCLUSION

Current classifications remain necessary but insufficient alone; treatment selection increasingly requires a dynamic, multi-axial framework reassessed at defined checkpoints rather than fixed at diagnosis.