(Urotoday.com) The 2023 GU ASCO annual meeting included an oral abstract prostate cancer session, featuring a discussion presentation by Dr. Tyler Seibert recapping two abstracts including “Patient-level data meta-analysis of a multi-modal artificial intelligence prognostic biomarker in high-risk prostate cancer: Results from six NRG/RTOG phase III randomized trials” presented by Dr. Daniel Spratt and “FORMULA-509: A multicenter randomized trial of post-operative salvage radiotherapy and 6 months of GnRH agonist with or without abiraterone acetate/prednisone and apalutamide post-radical prostatectomy” presented by Dr. Paul Nguyen.
First, Dr. Seibert discussed Spratt et al., specifically asking the question: Does the multi-modal artificial intelligence model stratify risk of distant metastasis and prostate cancer specific mortality among patients with high-risk prostate cancer undergoing radiotherapy and ADT? Dr. Seibert notes that there are several important characteristics from this cohort, including (i) a sample size of >1,000 from six prospective, cooperative-group trials, (ii) most of the patients had high Gleason score (>=8), and (iii) 20% of the population was black. In this study, on multivariable analysis, the multi-modal artificial intelligence model, adjusting for either age, PSA, Gleason score, T-stage, or number of high-risk features, was the only variable significantly associated with distant metastasis. Time dependent area under the curve was highest for the multi-modal artificial intelligence biomarker for both 5-year distant metastasis (0.71), compared to PSA (0.56), Gleason score (0.61), T-stage (0.63), or number of high-risk features (0.64), and for 15-year prostate cancer-specific mortality (0.73):
Furthermore, the estimated 10-year distant metastasis rates for multi-modal artificial intelligence quartile 1 vs 4 were 8% vs 31%:
Dr. Seibert postulated as to what would the absolute risks look like in a modern cohort outside of clinical trials? And, how accurate can these predictions be made? Additional downstream effects of this data may be that we are able to, with artificial intelligence, eliminate the need for central review:
- How robust is the multi-modal artificial intelligence model to inaccuracies in the clinical data inputs?
- PSA dynamics?
- Gleason score (with no central review)?
- T-stage (with no central review)?
With regards to the imagery data component, will all slides need to be submitted or only the ones with the worst pathology? Will MRI-guided biopsy sway multi-modal artificial intelligence results? Ultimately, Dr. Seibert emphasized that quantitative, objective, reproducible biomarkers will help us move away from over dependence on subjective expertise for diagnostics. Dr. Seibert’s take-away points from this presentation by Dr. Spratt are as follows:
- The multi-modal artificial intelligence refines prognosis within high-risk prostate cancer, which can be used to improve informed decision-making with each patient
- Do the multi-modal artificial intelligence model and Decipher genomic classifier provide redundant versus complementary information?
- Can we learn anything form pathology artificial intelligence features/phenotypes that could guide specific therapy choices?
Second, Dr. Seibert discussed Nguyen et al., specifically asking the question: how should we manage persistent disease after radical prostatectomy in the setting of high-risk features? Furthermore, who benefits from escalation with abiraterone acetate + prednisone + apalutamide? In this study, the overall cohort did not have a significant benefit by pre-specified threshold, and the pre-specified (and stratified) subgroup with PSA <= 0.5 ng/mL also did not benefit. Thus, Dr. Seibert notes that a single high-risk feature is not enough to prompt escalation. However, the pre-specified (and stratified) subgroup with a PSA > 0.5 ng/mL did benefit: for metastasis-free survival [HR 0.32 (90% CI 0.13-0.84), p=0.02 (2-sided); 3-year metastasis-free survival 66.1% bicalutamide vs. 84.3% abiraterone acetate + prednisone + apalutamide]:
Dr. Seibert questions whether a PSA > 0.5 ng/mL merely is an index of greater disease burden? Or is it a surrogate for more aggressive disease? If these patients had been treated before reaching a PSA of 0.51 ng/mL would they still have benefited from abiraterone acetate + prednisone + apalutamide. Importantly, the pre-specified (and stratified) subgroup of pN1 patients did not benefit from treatment intensification:
Perhaps, secondary to the large confidence intervals, the subgroup was underpowered. However, Dr. Seibert notes that early treatment failure was very common regardless of arm, thus this may suggest that 6 months of systemic therapy is too short for pN1 (even with abiraterone acetate + prednisone + apalutamide). If pN1 status really is not helpful for decision-making, we may need to examine the value of pelvic lymph node dissection (and optimal extent of pelvic lymph node dissection) in patients where post-op pelvic radiotherapy is likely needed. Dr. Seibert’s take-away points from this presentation by Dr. Nguyen are as follows:
- Escalation by 24 months has the strongest evidence today, specifically from the RADICALS-HD trial, with more than 1,500 men with 10-year follow-up
- Intensification for six months is a very compelling concept, and we await long-term follow-up of FORMULA-509 and the pending PROSTATE-IQ trial
Presented by: Tyler M. Seibert, MD, PhD University of California San Diego, San Diego, CA
Written by: Zachary Klaassen, MD, MSc – Urologic Oncologist, Assistant Professor of Urology, Georgia Cancer Center, Augusta University/Medical College of Georgia, @zklaassen_md on Twitter during the 2023 Genitourinary (GU) American Society of Clinical Oncology (ASCO) Annual Meeting, San Francisco, Thurs, Feb 16 – Sat, Feb 18, 2023.