Improving Prostate Cancer Prediction During Active Surveillance
Summary
- Dr. Sanoj Punnen and his research team at Desai Sethi Urology Institute are improving risk assessment for prostate cancer patients under active surveillance.
- The traditional way of assessing prostate cancer involved biopsies, which weren’t as accurate as doctors wanted.
- Dr. Punnen thinks artificial intelligence has promise in terms of improving MRI precision in active surveillance prostate cancer patients.
At the American Urologic Association (AUA) 2025 annual meeting, Sanoj Punnen, M.D., summarized the decade-long Miami Active Surveillance Trial (MAST) clinical trial and outlined how researchers are improving risk assessment for prostate cancer patients under active surveillance.
“My colleagues at AUA who manage prostate cancer patients on active surveillance understand how lingering uncertainty about whether a cancer is progressing and the need for serial biopsies and imaging can cause stress and anxiety for many patients,” said Dr. Punnen, the principal investigator for the MAST trial and professor of urologic oncology and vice chair of research at Desai Sethi Urology Institute (DSUI) at the University of Miami Miller School of Medicine. “Our research is helping to reduce this uncertainty by using a number of novel modalities to improve cancer risk assessment and enhance the precision of predicting cancer progression. We hope these improvements will allow us to lessen the burdens of surveillance in men who are unlikely to progress and intensify them in men with more riskier tumors to allow more timely detection and treatment.”
Few urological or cancer centers enroll prostate cancer patients in active surveillance clinical trials, according to Dipen J. Parekh, M.D., DSUI’s founding director and chief operating officer at UHealth—the University of Miami Health System.
“In this National Cancer Institute-funded clinical trial, we enrolled 200 men on a prospective protocol to validate the tools and technology being used in active surveillance today, such as prostate MRI and genomic testing, and have reported our outcomes in numerous papers,” Dr. Parekh said. “We also are taking this important research to the next level. For example, we are developing and validating AI tools to enhance MRI, pathology and biomarker detection to improve standard of care.”
The Evolution of Active Prostate Surveillance
Use of prostate MRI and genomic markers in active surveillance was in its infancy when Dr. Punnen and colleagues started the MAST study in 2014.
“Prior to that, the way we diagnosed prostate cancer was with randomly guided biopsies of the prostate,” Dr. Punnen said. “We would do an ultrasound to identify where the prostate was and then in a grid-like manner sample it in 12 different areas. For the most part, if a patient had a sizable bad tumor, you would find it that way. But one could miss a small, high-grade tumor with that approach.”
Surveilling the cancer meant doing another biopsy. Even then, there was always the chance the clinician may have missed the area of the prostate harboring a worse cancer.
“MRI was a real game changer in that we could look at the whole prostate and had a better sense of what was and what was not in the prostate,” Dr. Punnen said.
MRI, Genetic Testing and Prostate Cancer
The MAST trial explored whether annual MRI might allow patients and clinicians enough confidence to eliminate serial biopsies or prompt them to do biopsies, if the MRI signaled worsening cancer.
In a MAST-related study published in April 2025, the authors reported that MRI helped to predict which men would progress and require treatment. The findings suggest that men who came into the trial with positive findings on MRI were more likely to progress than those with a negative MRI. If the MRI got worse, patients were more likely to progress.
But MRI was far from perfect in its predictions, with significant numbers of false positives and false negatives. This suggests the current technology, alone, is not enough.
Another recently published study looking at genetic testing identified limitations to that approach, as well.
“Typically, the way these genetic tests work, you would send the worst core (the highest grade or highest volume) and a genetic testing company would look at the RNA and genetic profile of the tumor to help determine whether it is likely to progress or not,” Dr. Punnen explained. “We found, however, if you didn’t sample the worst core, you were not going to get accurate genetic information. The testing gives clinicians genetic information on the area they sampled and not about the tumor that may have been missed.”
The Future of Prostate Cancer Prognosis
Dr. Punnen says active surveillance prostate cancer patients fall into one of three buckets, with each bucket representing about one-third of patients. One bucket includes men likely to progress. The variable is timing and how long they can hold off on treatment. The second bucket includes men whose prognosis is unknown. The third bucket includes men who are unlikely to progress.
“With men in the third bucket, you try to spare them as much burden from the watching process as you can. The problem, however, is that we still don’t have good tools to reliably put patients into their respective buckets,” Dr. Punnen said. “We hope that some of the developments we are working on will help towards enhancing our precision in doing this.”
Dr. Punnen thinks artificial intelligence (AI) has promise in terms of improving MRI precision for active surveillance patients.
“What a machine gets with a pixel-by-pixel reading is much better than the naked eye. That’s where we’re doing a lot of work now,” he said.
The MAST trial opened the door to providing men considering active surveillance with a more personalized follow-up strategy based on the risk of progression. DSUI researchers are studying how to improve MRI imaging and the precision of genomic testing, pathologic interpretation and risk assessment, as well as biomarker development and validation in prostate cancer. AI could play a role in enhancing the current standard of care.
Dr. Punnen and colleagues are focused on validating these elements in larger studies, with more institutions involved.
“MAST is a single-center study of 200 men. How well it applies to a man living in California or some other country, we don’t really know,” Dr. Punnen said. “The foundation for this research is there. We are on course to evaluate AI integration to ultimately give each man and clinician clarity on the optimal and least invasive way to approach active surveillance.”
Tags: active surveillance, AI, artificial intelligence, Desai Sethi Urology Institute, Dr. Dipen J. Parekh, Dr. Sanoj Punnen, genetics, prostate cancer, technology