NON-MUSCLE-INVASIVE bladder cancer (NMIBC) makes up around 70% of all bladder cancers, with high-risk NMIBC (HR-NMIBC) being especially challenging to treat. Standard care involves transurethral resection of bladder tumor (TURBT) followed by intravesical bacillus Calmette-Guérin (BCG) therapy, or radical cystectomy for the highest-risk cases. Despite treatment, HR-NMIBC has a five-year recurrence rate of about 50% and progression rate of 20%. BCG intolerance leads to discontinuation in some patients, complicating management further.
A major challenge in HR-NMIBC is the lack of reliable tools to predict recurrence or progression after TURBT and BCG therapy. Recent advances in artificial intelligence (AI) offer new possibilities. AI has shown promise in other cancers, quantifying morphological features to predict treatment outcomes. Building on this, a recent multicenter study developed and validated the computational histology AI (CHAI) platform, a deep learning-based AI model designed to predict outcomes in HR-NMIBC.
The CHAI platform analyses digital whole-slide images (WSIs) from pre-treatment TURBT specimens of BCG-naïve HR-NMIBC patients. It identifies individuals at high or low risk for high-grade recurrence, progression, BCG-unresponsive disease (BUD), and the need for cystectomy. The model demonstrated strong predictive capabilities, identifying patients at 2.1, 3.9, 2.3, and 3.4-fold higher risk of recurrence, progression, BUD, and cystectomy, respectively.
Notably, AI-driven models outperformed traditional clinical risk calculators like the EORTC and EAU models, particularly in identifying patients at high risk after BCG therapy. The CHAI platform integrates histologic data with clinical features like multifocality and T1 disease for more accurate risk stratification.
These AI-based assays offer a promising approach to tailor treatment for HR-NMIBC patients. By helping clinicians select the most appropriate therapies, they can also guide clinical trial enrolment and surveillance regimens. Identifying patients less likely to respond to BCG therapy alone could lead to more effective alternative treatment options. While promising, further prospective clinical trials are needed to validate the platform’s real-world applicability.
Katie Wright, EMJ
Reference
Lotan Y et al. Predicting response to intravesical bacillus calmette-guérin in high-risk nonmuscle-invasive bladder cancer using an artificial intelligence-powered pathology assay: development and validation in an international 12-center cohort. J Urol. 2025;213(2):192-204.