PROGRxN-BCa AI Model Shows Broad Accuracy in NMIBC Prediction

The PROGRxN-BCa study, led by Jethro C.C. Kwong, a urology resident at the University of Toronto, showcases a major advancement in predicting disease progression in non–muscle-invasive bladder cancer (NMIBC) through artificial intelligence (AI). Using the largest NMIBC dataset ever assembled of 12,659 patients, the AI model was trained on 3,324 individuals from 4 Canadian hospitals and externally validated on an additional 9,335 patients across over 30 institutions in Canada, the US, and Europe.

What sets PROGRxN-BCa apart is not just its scale, but its broad applicability. In head-to-head comparisons, the model outperformed the European Association of Urology (EAU) risk calculator by approximately 10% in predicting progression to muscle-invasive or metastatic disease. The study underscores how an AI tool trained on real-world, readily available clinical and pathological data can yield more accurate and actionable risk predictions—especially for the complex and heterogeneous intermediate-risk group, where traditional tools often fall short.

“We chose to take all comers of patients with NMIBC, and we looked at the performance of our model in many different ways, and we have shown that it works well for old, young, male, female—their socioeconomic status—it did not matter, the model performed well,” Kwong explains.

Critically, the model's performance held steady regardless of whether patients received what is considered “guideline-concordant care,” such as repeat transurethral resection (TURBT) for T1 tumors or appropriately administered BCG therapy. “There are many reasons why a patient may not receive guideline-concordant care—comorbidities, treatment access, or even institutional practice variation,” Kwong notes. “What we were able to show is that the model performs well for both groups.”

By demonstrating consistent predictive performance across demographic and clinical care variations, PROGRxN-BCa offers a generalizable and scalable approach to NMIBC risk stratification. It holds promise for broader clinical use, informing more personalized treatment decisions and potentially reshaping future bladder cancer care guidelines.