Researchers have developed a novel artificial intelligence (AI) model that significantly improves prediction of progression risk in patients with non-muscle invasive bladder cancer (NMIBC), particularly for those in the challenging intermediate-risk category.
The AI-based tool, named PROGRxN-BCa, was trained on the largest NMIBC cohort to date, comprising 12,659 patients across multiple institutions in North America and Europe. The model demonstrated approximately 10% better performance than the current guideline-endorsed European Association of Urology (EAU) risk calculator.
"This was an AI model that was trained in the largest NMIBC cohort in the world, with over 12,000 patients, and it is actually over 100 times larger than the median cohort size of prior AI studies," explained Jethro C.C. Kwong, a urology resident from the University of Toronto who led the research.
Development and Validation Process
The research team trained PROGRxN-BCa using data from 3,324 patients treated at four academic and community hospitals in Canada. The model incorporates 14 clinical and pathological features that are routinely available in day-to-day practice, requiring no specialized histopathological slides or biomarkers.
External validation was then conducted on a separate cohort of 9,335 patients from over 30 institutions across Canada, the United States, and Europe. This extensive validation demonstrated consistent superior performance regardless of adherence to guideline-concordant care.
"We built this model, and then we externally validated this on over 9000 patients across Canada, US, and Europe, and there were over 30 institutions involved for this," Kwong noted.
Addressing Critical Unmet Needs
The development of PROGRxN-BCa addresses significant limitations in current risk stratification tools for NMIBC. Existing calculators have shown poor performance in external validation and often fail to reflect contemporary clinical practice. This is particularly problematic given the recent influx of new treatment options for NMIBC patients.
One of the most significant advantages of the new model is its ability to effectively sub-stratify the heterogeneous intermediate-risk group into distinct risk tertiles. This capability allows clinicians to better identify patients with varying progression probabilities—a substantial improvement over current methods that struggle to differentiate risk within this category.
"We were able to show that our AI model outperforms the current guideline-endorsed tool, generally by about 10%. It is also quite helpful. What we have shown is kind of a practical application of this tool in trying to sub-stratify intermediate-risk normal cell-based bladder cancer," Kwong emphasized.
Clinical Implications
The improved risk stratification offered by PROGRxN-BCa has significant implications for clinical decision-making. By more accurately identifying patients at higher risk of progression to muscle-invasive or metastatic disease, clinicians can tailor treatment approaches more precisely.
For intermediate-risk patients, who represent a particularly heterogeneous group, the model's ability to sub-stratify into distinct risk categories could help determine which patients might benefit from more aggressive surveillance or treatment versus those who could safely undergo less intensive management.
The model's reliance on readily available clinical and pathological features also makes it practical for implementation in various clinical settings without requiring additional specialized testing or resources.
Future Directions
While the initial results are promising, further research will likely focus on prospective validation and implementation studies to determine how the model performs in real-time clinical decision-making. Additionally, integration of the AI tool into electronic health records and clinical workflows will be essential for widespread adoption.
The development of PROGRxN-BCa represents a significant step forward in the application of artificial intelligence to improve cancer risk prediction and potentially patient outcomes through more personalized treatment approaches.
