Development and Validation of a Multimodal Artificial Intelligence-derived Digital Pathology-based Biomarker Predicting Metastasis Among Patients with Biochemical Recurrence After Radical Prostatectomy in NRG/RTOG Trials

Todd M Morgan; Yi Ren; Siyi Tang; Wouter Zwerink; Emmalyn Chen; Akinori Mitani; Huei-Chung Huang; Jeffry P Simko; Sandy DeVries; Alan Pollack; Derek Wilke; André-Guy Martin; Alexander G Balogh; Jeff M Michalski; Michael J Greenberg; Jason A Efstathiou; Jean-Paul Bahary; Ashley E Ross; Andre Esteva; Trevor J Royce; Paul L Nguyen; Karen E Hoffman; Howard M Sandler; Phuoc T Tran; Stephanie L Pugh; Felix Y Feng; Daniel E Spratt

doi:10.1016/j.eururo.2025.12.007

Development and Validation of a Multimodal Artificial Intelligence-derived Digital Pathology-based Biomarker Predicting Metastasis Among Patients with Biochemical Recurrence After Radical Prostatectomy in NRG/RTOG Trials

Eur Urol. 2025 Dec 22:S0302-2838(25)04859-6. doi: 10.1016/j.eururo.2025.12.007. Online ahead of print.

Authors

Todd M Morgan¹, Yi Ren², Siyi Tang², Wouter Zwerink², Emmalyn Chen², Akinori Mitani², Huei-Chung Huang², Jeffry P Simko³, Sandy DeVries³, Alan Pollack⁴, Derek Wilke⁵, André-Guy Martin⁶, Alexander G Balogh⁷, Jeff M Michalski⁸, Michael J Greenberg⁹, Jason A Efstathiou¹⁰, Jean-Paul Bahary¹¹, Ashley E Ross¹², Andre Esteva², Trevor J Royce², Paul L Nguyen¹³, Karen E Hoffman¹⁴, Howard M Sandler¹⁵, Phuoc T Tran¹⁶, Stephanie L Pugh¹⁷, Felix Y Feng¹⁸, Daniel E Spratt¹⁹

Affiliations

¹ University of Michigan Comprehensive Cancer Center, Ann Arbor, MI, USA. Electronic address: tomorgan@med.umich.edu.
² Artera Inc., Los Altos, CA, USA.
³ NRG Oncology Biospecimen Bank, San Francisco, CA, USA; UCSF Medical Center-Mission Bay, San Francisco, CA, USA.
⁴ University of Miami Miller School of Medicine, Miami, FL, USA.
⁵ Capital Health - Nova Scotia Cancer Centre, Halifax, NS, Canada.
⁶ CHU de Québec- Université Laval, Quebec, QC, Canada.
⁷ Tom Baker Cancer Centre, Calgary, AB, Canada.
⁸ Washington University School of Medicine, St Louis, MO, USA.
⁹ Thomas Jefferson University Hospital, Philadelphia, PA, USA.
¹⁰ Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA.
¹¹ Centre Hospitalier de l'Université de Montréal-Notre Dame, Montreal, QC, Canada.
¹² Northwestern University, Chicago, IL, USA.
¹³ Dana-Farber Cancer Institute, Boston, MA, USA.
¹⁴ MD Anderson Cancer Center, Houston, TX, USA.
¹⁵ Cedars-Sinai Medical Center, Los Angeles, CA, USA.
¹⁶ University of Maryland Medical System, Baltimore, MD, USA.
¹⁷ NRG Oncology Statistics and Data Management Center, Philadelphia, PA, USA; American College of Radiology, Philadelphia, PA, USA.
¹⁸ UCSF Medical Center-Mission Bay, San Francisco, CA, USA.
¹⁹ University Hospitals at Case Western Reserve University, Cleveland, OH, USA. Electronic address: daniel.spratt@uhhospitals.org.

Abstract

Background and objective: Biochemical recurrence (BCR) after radical prostatectomy (RP) is a heterogeneous disease state in prostate cancer with multiple treatment options. Improved risk stratification could enable more personalized decision-making. We developed and validated a digital pathology-based multimodal artificial intelligence (MMAI) model to predict outcomes in post-RP BCR patients undergoing salvage therapy.

Methods: An MMAI model was trained to predict distant metastasis (DM) using prostate histopathology image features and clinical variables (pathologic grade group, pathologic T stage, prostate-specific antigen level before salvage radiotherapy [SRT], age, and surgical margin). The locked model was validated in 533 patients from NRG/RTOG 9601 and 0534 treated with SRT ± hormone therapy (HT), using Cox regression and time-dependent area under the receiver operating characteristic curve.

Key findings and limitations: With a median follow-up of 9.3 yrs, MMAI score was significantly associated with DM (subdistribution hazard ratio = 2.17 per standard deviation [95% confidence interval 1.65-2.85]; p < 0.001) and remained independently prognostic after adjusting for clinical variables and treatment. The 10-yr time-dependent area under the receiver operating characteristic curve for MMAI was 0.74 compared with 0.68 for a clinical nomogram. Binary risk categorization demonstrated higher 10-yr DM incidence in the MMAI high-risk (25%) than in the low-risk (8.8%) group. The absolute reduction in 10-yr DM incidence with HT plus SRT versus SRT alone was 21% in the high-risk group versus 2.5% in the low-risk group. Limitations include the use of archived trial cohorts.

Conclusions and clinical implications: The post-RP MMAI model provides individualized risk estimates after SRT ± HT and may support shared decision-making about salvage treatment. External and prospective validation are ongoing.

Keywords: Artificial intelligence; Digital pathology; Prognostic model; Prostate cancer; Radical prostatectomy; Salvage therapy.

Abstract

Grants and funding