An Artificial Intelligence Predictive Biomarker for Response to BCG vs Gemcitabine-Docetaxel for High Grade NMIBC - Vignesh Packiam
July 24, 2025
Vignesh Packiam discusses an AI biomarker developed for nonmuscle invasive bladder cancer that addresses the critical challenge of risk stratification in this highly heterogeneous disease. The artificial intelligence test analyzes digitized H&E slides from the index TURBT, examining over 600 microscopic features including immune signatures, stromal features, and nuclear characteristics to predict patient response to treatment. With new FDA-approved therapies emerging beyond the decades-old BCG standard, Dr. Packiam emphasizes that accurate biomarkers are more crucial than ever for selecting optimal treatments and stratifying patients into likely responders versus non-responders.
Biographies:
Vignesh Packiam, MD, Director of Clinical and Translational Research in Urologic Oncology, Associate Professor of Surgery, Rutgers Cancer Institute of New Jersey, RWJ Barnabas Health, New Brunswick, NJ
Ashish Kamat, MD, MBBS, Professor of Urology and Wayne B. Duddleston Professor of Cancer Research, University of Texas, MD Anderson Cancer Center, Houston, TX
Biographies:
Vignesh Packiam, MD, Director of Clinical and Translational Research in Urologic Oncology, Associate Professor of Surgery, Rutgers Cancer Institute of New Jersey, RWJ Barnabas Health, New Brunswick, NJ
Ashish Kamat, MD, MBBS, Professor of Urology and Wayne B. Duddleston Professor of Cancer Research, University of Texas, MD Anderson Cancer Center, Houston, TX
Related Content:
AUA 2025: An Artificial Intelligence-powered Predictive Biomarker for Response to Intravesical BCG versus Gemcitabine-Docetaxel for High-Grade Non-Muscle Invasive Bladder Cancer
Computational Histology AI's Potential in Refining NMIBC Prognostication - Bryn Launer
SUO 2024: An Artificial Intelligence Powered Model to Predict Response to Intravesical BCG Versus Gemcitabine-Docetaxel for High Risk NMIBC
AUA 2025: An Artificial Intelligence-powered Predictive Biomarker for Response to Intravesical BCG versus Gemcitabine-Docetaxel for High-Grade Non-Muscle Invasive Bladder Cancer
Computational Histology AI's Potential in Refining NMIBC Prognostication - Bryn Launer
SUO 2024: An Artificial Intelligence Powered Model to Predict Response to Intravesical BCG Versus Gemcitabine-Docetaxel for High Risk NMIBC
Read the Full Video Transcript
Vignesh Packiam: Appreciate you having me on this topic. This is something we've been working on for the last few years, and it's finally yielding some exciting results that I'm excited to discuss.
So this project is looking at an AI biomarker that originally was developed on a cohort of patients that received BCG. And more recently, we assessed it in a new cohort of patients that either got BCG or Gem/Doce. So as a background, we really do need good biomarkers for nonmuscle invasive bladder cancer. We know that NMIBC is highly heterogeneous.
There's so much variability based on clinical pathologic features that it's hard to really extract purely from a pathology report or from demographic variables. And over time, there have been various different risk stratification guidelines that have been developed by the AUA, EAU, NCCN, IBCG, and others.
And even between those guidelines, there's a lot of variability with how to risk stratify patients purely based on clinical pathologic features. Historically, our treatments have been relatively homogeneous for nonmuscle invasive bladder cancer. We've been using BCG for decades. And that's stayed the standard of care for high-grade disease for a very long time.
But recently, we've had more new therapies that have been developed. Some are off-label, some are FDA approved, some are emerging. And now that we have more treatment options, I think biomarkers are needed more than ever to be able to accurately choose which one is best for which patient.
So talking about this assay, this is an artificial intelligence test that generates biomarkers using digitized H&E slides from an index TURBT from the beginning of the diagnosis for bladder cancer. So what it does is it takes a representative slide from the TURBT, gets the digitized image. The image is magnified at 40x.
Microscopically, the AI algorithm looks for features that are associated with risk of recurrence or progression after BCG. Those include immune signatures, stromal features, nuclear features as well. I think there's over 600 that are analyzed by the algorithm. And then it takes all of these, and it helps stratify the patients into those that are likely to respond or unlikely to respond.
Vignesh Packiam: Appreciate you having me on this topic. This is something we've been working on for the last few years, and it's finally yielding some exciting results that I'm excited to discuss.
So this project is looking at an AI biomarker that originally was developed on a cohort of patients that received BCG. And more recently, we assessed it in a new cohort of patients that either got BCG or Gem/Doce. So as a background, we really do need good biomarkers for nonmuscle invasive bladder cancer. We know that NMIBC is highly heterogeneous.
There's so much variability based on clinical pathologic features that it's hard to really extract purely from a pathology report or from demographic variables. And over time, there have been various different risk stratification guidelines that have been developed by the AUA, EAU, NCCN, IBCG, and others.
And even between those guidelines, there's a lot of variability with how to risk stratify patients purely based on clinical pathologic features. Historically, our treatments have been relatively homogeneous for nonmuscle invasive bladder cancer. We've been using BCG for decades. And that's stayed the standard of care for high-grade disease for a very long time.
But recently, we've had more new therapies that have been developed. Some are off-label, some are FDA approved, some are emerging. And now that we have more treatment options, I think biomarkers are needed more than ever to be able to accurately choose which one is best for which patient.
So talking about this assay, this is an artificial intelligence test that generates biomarkers using digitized H&E slides from an index TURBT from the beginning of the diagnosis for bladder cancer. So what it does is it takes a representative slide from the TURBT, gets the digitized image. The image is magnified at 40x.
Microscopically, the AI algorithm looks for features that are associated with risk of recurrence or progression after BCG. Those include immune signatures, stromal features, nuclear features as well. I think there's over 600 that are analyzed by the algorithm. And then it takes all of these, and it helps stratify the patients into those that are likely to respond or unlikely to respond.