(UroToday.com) The 2025 American Society of Clinical Oncology (ASCO) Genitourinary (GU) Annual Symposium held in San Francisco, CA was host to a prostate cancer poster session. Dr. Matthew Cooperberg presented a study using a validated pathology-based multimodal artificial intelligence (MMAI) biomarker to predict long-term outcomes following radical prostatectomy (RP).
RP improves metastasis-free and overall survival in higher risk localized prostate cancer patients,1 but outcomes vary dramatically with baseline risk characteristics. Predictive biomarkers for improved risk stratification and treatment-decision making after RP require tissue destruction and are subject to intratumor variability. MMAI models utilizing image analysis of standard clinical pathology biopsy slides have previously been validated as prognostic, as well as predictive, biomarkers of response to intensified treatment among patients undergoing radiation therapy2 and have been endorsed in the NCCN guidelines. The study investigators sought to determine the utility of the MMAI model originally developed using whole slide images from prostate biopsies for predicting outcomes after RP using tissue microarray (TMA) specimens with rich clinical annotation and long-term follow-up.
A TMA from 424 low- to intermediate-risk RP cases with long-term follow-up from the University of California, San Francisco was generated.
TMA was scanned at high resolution, and MMAI (scored 0-1) and Cancer of the Prostate Risk Assessment pre- and post-surgical (CAPRA [0-10], CAPRA-S [0-12]) scores were calculated.
The study outcomes assessed were:
- Adverse pathology (i.e., pT3a or worse and/or grade group ≥3)
- Recurrence (two PSA tests ≥0.2 ng ml or any second treatment)
- Metastasis after RP
Multivariable logistic regression (for adverse pathology) and Cox proportional hazards regression modeling (for recurrence and metastasis) were performed, controlling for CAPRA-S.
Odds ratios (OR) and hazard ratios (HR) were determined per 0.1 increase in MMAI score.
The baseline characteristics of patients forming the TMA cohort are summarized below. The median (IQR) follow-up was 13.2 (7.8-18.3) years. By CAPRA-S, 273 (66%), 114 (28%), and 24 (6%) were low (0-2), intermediate (3-5), and high (≥6) risk, respectively.
MMAI scores were successfully generated for 414 (98%) of cases from the TMA. The median (IQR) MMAI risk score was 0.25 (0.18-0.33). The correlation between CAPRA and MMAl scores was statistically significant but weak (r=0.33, p <0.01).
On logistic regression adjusting for CAPRA, the MMAI score was significantly associated with the risk of adverse pathology (OR: 1.02, 95% CI: 1.00–1.05).
Recurrence-free and metastasis-free survivals were 74% and 96% at 10 years in the cohort. On unadjusted Cox regression, MMAl risk score was significantly associated with risk of both recurrence (HR 1.04. 95% CI 1.02-1.06) and metastasis (HR 1.05, 95% Cl 1.02-1.07), but was not significant after adjustment for CAPRA-S, which includes post-surgical factors.
Dr. Cooperberg acknowledged that important limitations were present:
- The MMAI model was not developed for use with TMA, and sampling variability due to the capture of less than a full prostatectomy slide image may affect the results
- The study cohort was from a single academic institution with a relatively homogeneous population
- This was a relatively low-risk cohort; many would be eligible for active surveillance in the modern era
- This analysis lacked data for modern imaging (e.g. mpMRI, PSMA PET) or additional diagnostic tools
Dr. Cooperberg concluded as follows:
- The MMAI score improved over the CAPRA score for predicting adverse pathology at radical prostatectomy. It was also found to be associated with recurrence and metastasis outcomes, following adjustment for the CAPRA-S, but did not reach statistical significance in this relatively low risk cohort
- Compared to other markers, MMAI is non-destructive and less subject to intratumoral variation. He noted that it will hopefully offer faster turnaround and lower cost as well.
Presented by: Matthew R. Cooperberg. MD, MPH, Professor of Urology and Epidemiology & Biostatistics and Helen Diller Family Chair in Urology at the University of California, San Francisco, CA
Written by: Rashid K. Sayyid, MD, MSc – Robotic Urologic Oncology Fellow at The University of Southern California, @rksayyid on Twitter during the 2025 Genitourinary (GU) American Society of Clinical Oncology (ASCO) Annual Meeting, San Francisco, CA, Thurs, Feb 13 – Sat, Feb 15, 2025.
References:
- Hamdy FC, Donovan JL, Lane JA, et al. Fifteen-Year Outcomes after Monitoring, Surgery, or Radiotherapy for Prostate Cancer. N Engl J Med. 2023; 388(17):1547-58.
- Spratt DE, Tang S, Sun Y, et al. Artificial Intelligence Predictive Model for Hormone Therapy Use in Prostate Cancer. N Engl J Med Evid. 2023; 2(8).