JU INSIGHT Multimodal Biomarkers That Predict Gleason Pattern 4: Potential Impact for Active Surveillance

By: D. M. Berman, MD, PhD, Queen’s University Cancer Research Institute, Kingston, Ontario, Canada; A. Y. Lee, PhD, Ontario Institute for Cancer Research, Toronto, Canada; R. Lesurf, PhD, Ontario Institute for Cancer Research, Toronto, Canada; P. G. Patel, PhD, Queen’s University Cancer Research Institute, Kingston, Ontario, Canada; W. Ebrahimizadeh, PhD, McGill University Health Centre, Montréal, Québec, Canada; J. Bayani, PhD, Ontario Institute for Cancer Research, Toronto, Canada, University of Toronto, Ontario, Canada; L. A. Lee, MSc, PMP, Ontario Institute for Cancer Research, Toronto, Canada; N. Boufaied, PhD, McGill University Health Centre, Montréal, Québec, Canada; S. Selvarajah, PhD, Queen’s University, Kingston, Ontario, Canada; T. Jamaspishvili, MD, PhD, Queen’s University Cancer Research Institute, Kingston, Ontario, Canada; K.-P. Guérard, MSc, McGill University Health Centre, Montréal, Québec, Canada; D. Dion, BA, Ontario Institute for Cancer Research, Toronto, Canada; A. Kawashima, MD, PhD, Queen’s University Cancer Research Institute, Kingston, Ontario, Canada; G. M. Clarke, PhD, Ontario Institute for Cancer Research, Toronto, Canada; N. How, MD, Queen’s University Cancer Research Institute, Kingston, Ontario, Canada, Hamilton Health Sciences, Ontario, Canada; C. L. Jackson, PhD, Queen’s University Cancer Research Institute, Kingston, Ontario, Canada; E. Scarlata, DVM, PhD, McGill University Health Centre, Montréal, Québec, Canada; K. Siddiqui, B.Eng, PhD, Schulich School of Medicine and Dentistry, University of Western Ontario, London, Canada; J. B. A. Okello, PhD, Queen’s University Cancer Research Institute, Kingston, Ontario, Canada; A. G. Aprikian, MD, McGill University Health Centre, Montréal, Québec, Canada; M. Moussa, MBBCh, Schulich School of Medicine and Dentistry, University of Western Ontario, London, Canada, London Health Sciences Centre, Ontario, Canada; A. Finelli, MD, MSc, Princess Margaret Cancer Centre. Toronto, Ontario, Canada, University of Toronto, Ontario, Canada; J. Chin, MD, Schulich School of Medicine and Dentistry, University of Western Ontario, London, Canada, London Health Sciences Centre, Ontario, Canada; F. Brimo, MD, McGill University Health Centre, Montréal, Québec, Canada; G. Bauman, MD, Schulich School of Medicine and Dentistry, University of Western Ontario, London, Canada, London Health Sciences Centre, Ontario, Canada; A. Loblaw, MD, MSc, Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada, University of Toronto, Ontario, Canada; V. Venkateswaran, PhD, University of Toronto, Ontario, Canada; R. Buttyan, PhD, Vancouver Prostate Centre, British Columbia, Canada; S. Chevalier, PhD, McGill University Health Centre, Montréal, Québec, Canada; A. Thomson, PhD, McGill University Health Centre, Montréal, Québec, Canada; P. C. Park, MD, PhD, Queen’s University, Kingston, Ontario, Canada; D. R. Siemens, MD, PhD, Queen’s University Cancer Research Institute, Kingston, Ontario, Canada; J. Lapointe, MD, PhD, McGill University Health Centre, Montréal, Québec, Canada; P. C. Boutros, PhD, Ontario Institute for Cancer Research, Toronto, Canada, University of Toronto, Ontario, Canada; J. M. S. Bartlett, PhD, Ontario Institute for Cancer Research, Toronto, Canada, University of Toronto, Ontario, Canada, Edinburgh Cancer Research Centre, University of Edinburgh, United Kingdom | Posted on: 30 Aug 2023

Berman DM, Lee AY, Lesurf R, et al. Multimodal biomarkers that predict the presence of Gleason pattern 4: potential impact for active surveillance. J Urol. 2023;210(2):257-271.

Study Need and Importance

Due to biopsy sampling error and limited sensitivity of multiparametric MRI, diagnostic tools to identify cancers containing occult Gleason pattern 4 have limited accuracy and utility. Addressing this gap would identify men who could benefit from more intensive surveillance or treatment. We hypothesized that molecular biomarkers can detect the presence of Gleason pattern 4 and that combining DNA- and RNA-based features can improve biomarker accuracy.

What We Found

Using centrally reviewed radical prostatectomy pathology as a gold standard and grade group (GG) 1 vs GG ≥2 as an end point, we analyzed 467 molecular and clinical variables from 535 men. From GG ≥2 cancers, sampling included separate low- and high-grade samples. Applying 11,983 different machine learning strategies within cross-validation, we found that combining mRNA abundance, DNA copy number alteration, and DNA methylation features improved the area under the receiver operating characteristic curve (AUROC). Using a fully independent prostatectomy cohort, we validated 2 top-performing molecular classifiers, personalized risk stratification for patients with early prostate cancer (PRONTO)-e and PRONTO-m, with AUROCs of 0.79 and 0.82, respectively. Importantly, both classifiers were resistant to sampling error, with the ability to classify GG ≥2 cancers from their low-grade (GG1) regions, and both classifiers identified significantly more upgraded cases than a well-validated clinical risk calculator (Cancer of the Prostate Risk Assessment [CAPRA]). Using PRONTO-m as an example in a hypothetical cohort of 1,000 men on active surveillance, 86% of patients predicted to be negative would truly be GG1.

Limitations

Clinical validation in needle biopsies is required to demonstrate the utility and performance of these classifiers.

Interpretation for Patient Care

Upon clinical validation, classifiers with the performance characteristics of PRONTO-e and PRONTO-m could deintensify surveillance for men predicted to have true GG1 cancers (see Table). The same classifiers could prioritize men classified as GG ≥2 for more active investigation, identifying potentially harmful cancers in time for curative treatment.