CHAI biomarkers demonstrate prognostic utility for high-grade Ta NMIBC

Computational Histology Artificial Intelligence (CHAI) biomarkers demonstrated the ability to prognosticate outcomes for patients with high-grade (HG) Ta non–muscle-invasive bladder cancer (NMIBC), according to data published in European Urology.¹

According to Valar Labs, which developed the Vesta platform that uses CHAI biomarkers, this patient population is “frequently on the borderline of guideline-based risk stratification [and] may receive different treatment depending on which guideline their provider uses.”²

The CHAI platform was designed to assess risk of recurrence and progression in these patients by quantifying “histologic features, including tumor cell morphology and tumor microenvironment, from digital whole-slide images,” the authors explained.¹

Sam S. Chang, MD, MBA

“HG Ta tumors carry differing risks [that] current guidelines attempt to address, and clinical factors such as tumor focality and size can help delineate the true nature of these tumors. These clinical factors, however, are somewhat crude and can be misleading at times,” explained lead author Sam S. Chang, MD, MBA, Chief Surgical Officer of Vanderbilt Ingram Cancer Center, Nashville, Tennessee, in a news release from Valar.² “This study demonstrated that CHAI biomarkers can offer improved and additional risk stratification that can go beyond traditional risk factors to better prognosticate outcomes in this population.”

For the study, the investigators assessed the ability of the CHAI platform to predict the risk of recurrence and progression in a retrospective multicenter cohort of 269 patients with high-grade Ta disease. They then compared the performance of the platform to existing American Urological Association (AUA) and European Association of Urology (EAU) models. CHAI biomarker status was classified as either being present (indicative of higher risk) or absent. The median follow-up was 32 months (IQR, 14 to 52).

Data showed that EAU risk group was not significantly associated with high-grade recurrence-free survival (RFS) on either univariate analysis (high risk vs intermediate risk; HR, 1.47; 95% CI, 0.96-2.23; P = .074) or multivariable analysis adjusted for CHAI biomarker status. AUA high-risk was significantly associated with high-grade RFS on univariate analysis (HR, 1.93; 95% CI, 1.07-3.48; P = .029), but the association did not remain significant when accounting for CHAI biomarker status.

The CHAI biomarker, however, demonstrated the ability to stratify outcomes independent of these risk groups. The presence of the CHAI biomarker was found to be significantly associated with high-grade RFS on both univariate analysis (HR, 2.23; 95% CI, 1.45-3.44; P < .001) as well as multivariate analysis when adjusting for AUA and EAU risk groups or their individual components.

The authors also noted, “The continuous output for the CHAI biomarker had higher [areas under the curve] than the AUA and EAU risk groups for prediction of both endpoints along all time points.”

There was also an association observed between the AI pathology component of the CHAI biomarker (with tumor multifocality excluded) and high-grade RFS. On multivariate analysis, the AI pathology component alone provided prognostic information beyond AUA risk (HR, 2.21; 95% CI, 1.42-3.43; likelihood ratio P < .001) and EAU risk (HR, 2.19; 95% CI, 1.40-3.43; likelihood ratio P = .001) groups.

Likewise, neither EAU (HR, 1.32; 95% CI, 0.40 to 4.32; P = .7) nor AUA risk groups (HR, 1.39; 95% CI, 0.30-6.44; P = .7) were significantly associated with muscle-invasive bladder cancer (MIBC) progression-free survival (PFS). However, there was a significant association between the AI component of the CHAI biomarker and MIBC-PFS (HR, 4.55; 95% CI, 1.39-14.92; P = .012).

The association of CHAI biomarker with RFS and PFS was maintained even after the investigators excluded patients who had concurrent carcinoma in situ.

Siamak Daneshmand, MD

The authors noted that if the CHAI biomarker had been used to determine risk rather than the EAU and AUA stratifications, approximately 70% of patients deemed as having high risk would have been classified as being closer to intermediate risk. Further, 9.5% of patients designated as intermediate risk would have been classified as being at higher risk for recurrence and progression.

Senior author Siamak Daneshmand, MD, of the University of Southern California, concluded in the news release,² "The ability to derive clinically meaningful predictive and prognostic information from routinely-collected pathology is a paradigm shift for the risk stratification of bladder cancer patients."

REFERENCES

1. Chang SS, Launer B, Narayan V, et al. Computational histology artificial intelligence (CHAI) enhances risk stratification of high-grade Ta non–muscle-invasive bladder cancer in a multicenter cohort: comparison to current European Association of Urology and American Urological Association stratification schemes. Eur Urol. Published online June 12, 2025. doi:10.1016/j.eururo.2025.05.035

2. Valar Labs’ CHAI biomarkers validated in new study in European Urology for predicting outcomes of high grade Ta bladder cancer. News release. Valar Labs. June 13, 2025. Accessed June 13, 2025. https://www.businesswire.com/news/home/20250612200267/en/Valar-Labs-CHAI-Biomarkers-Validated-in-New-Study-in-European-Urology-for-Predicting-Outcomes-of-High-Grade-Ta-Bladder-Cancer