Risk Stratification and Selection of Management Strategy for Localized Prostate Cancer

Presented by:
Alice Yu
Search for other papers by Alice Yu in
Current site
Google Scholar
PubMed
Close
 MD, MPH
Full access

Localized prostate cancer presents a wide disease spectrum, ranging from indolent cases suitable for active surveillance to aggressive tumors requiring intensive multimodal treatment. Traditional risk stratification tools, such as the D’Amico risk categories and NCCN risk groups, have limitations because of their heterogeneity within each category. Incorporating novel risk stratification strategies, such as genomic classifiers and multimodal artificial intelligence assays, into clinical practice may help refine treatment decisions and optimize outcomes for patients with localized prostate cancer. However, it is important to recognize the limitations of these tools and to use them judiciously in the appropriate clinical contexts.

Novel risk stratification strategies, including tumor multigene molecular testing and multimodal artificial intelligence (MMAI) assays, are emerging as powerful tools to refine treatment decisions and optimize outcomes for patients with localized prostate cancer. During the NCCN 2024 Annual Conference, Alice Yu, MD, MPH, Assistant Member, Department of Genitourinary Oncology, Moffitt Cancer Center, and member of the NCCN Guidelines Panel for Prostate Cancer Early Detection, identified patient and disease characteristics that should be considered when recommending disease management approaches for patients with localized prostate cancer, and discussed strategies to refine risk stratification in appropriate patients.

Why Risk Stratify?

As Dr. Yu explained, risk stratification is essential for managing localized prostate cancer given the wide disease spectrum, ranging from indolent cases suitable for active surveillance to aggressive tumors requiring intensive multimodal treatment (Figure 1).

Figure 1.
Figure 1.

Localized prostate cancer: disease spectrum.

Abbreviations: ADT, androgen deprivation therapy; RP, radical prostatectomy.

Citation: Journal of the National Comprehensive Cancer Network 22, Supplement; 10.6004/jnccn.2024.5013

Although the D’Amico risk classification system has been widely used, she said, it exhibits heterogeneity within each risk group. A study by Klotz et al,1 for example, found that a small number of patients in active surveillance cohorts developed metastatic disease, even without upgrading to Gleason 7. Similarly, Reese et al2 reported that patients in the intermediate-risk category based on clinical stage had lower recurrence rates compared with those categorized by Gleason score 7 or prostate-specific antigen (PSA) level (10–20 ng/mL).

To address this heterogeneity, the NCCN risk classification further separates patients into categories. However, even the NCCN classification is not perfect, according to Dr. Yu, and there is still heterogeneity within each risk category. Because most risk groups are assigned based on clinical variables (clinical stage, Gleason score, and PSA level), the next step is to look beyond these variables and consider genomics.

According to Dr. Yu, newer risk classification schemes and tools, such as tumor multigene molecular testing and MMAI assays, can be used in addition to NCCN risk groups. However, “these tools should not be used for everyone and should only be considered when they have the potential to change management,” cautioned Dr. Yu.

Low- to Intermediate-Risk Prostate Cancer

Decipher, a genomic classifier (GC) assay, is a 22-gene RNA feature biomarker validated to predict metastasis in patients who have undergone radical prostatectomy and biopsy specimens. A study of 235 patients treated with prostatectomy or radiation from 7 referral centers found that Decipher was a better predictor for metastasis at 5 years after biopsy compared with traditional risk stratifiers such NCCN or CAPRA (Cancer of the Prostate Risk Assessment), with a C-index of 0.74 versus 0.66 and 0.6, respectively.3

Another study by Spratt et al4 classified patients into a novel clinical genomic risk classification system based on their GC assay. By combining the NCCN risk group and the GC risk classification, patients can be reclassified into a new risk category. For example, a patient with a favorable intermediate-risk and a low-risk GC score would be restratified into the low-risk group. This novel classifier performed well, with a C-index of 0.84 compared with 0.73 for traditional NCCN risk assessment.

The MUSIC cohort study, which included patients with low- and favorable intermediate-risk cancers who underwent a GC, found variability in GC scores within grade group 1 and intermediate-risk disease.5 Patients with a high-risk GC were more likely to undergo disease reclassification upon rebiopsy and treatment in the long run. However, Dr. Yu cautioned, the GC assay is a prognosticator, not a predictor.

“It remains unclear whether treating these patients earlier leads to clinically meaningful outcomes like improved metastasis-free survival, prostate cancer–specific survival, or overall survival,” she said. “In very-low-risk categories, the possibility of overtreatment should be considered.”

Two additional genomic assays that can be considered when deciding between surveillance and treatment for patients with low- to intermediate-risk prostate cancer are the 17-gene GPS (Genomic Prostate Score) assay and the 31-gene CCP (cell-cycle progression) assay (Figure 2).

Figure 2.
Figure 2.

Clinical scenarios where molecular biomarkers can be helpful.

Abbreviations: ADT, androgen deprivation therapy; CCP, cell-cycle progression; GC, genomic classifier; GPS, Genomic Prostate Score; LT-ADT, long-term androgen deprivation therapy; ST-ADT, short-term androgen deprivation therapy.

Citation: Journal of the National Comprehensive Cancer Network 22, Supplement; 10.6004/jnccn.2024.5013

The 17-gene GPS assay was developed using 3 cohorts: a prostatectomy cohort, a biopsy cohort, and a prospectively designed validation cohort from Cleveland.6 This assay is not intended for use in patients with high-risk disease, according to Dr. Yu. The assay was tested on biopsy specimens from patients with low to intermediate clinical risk; of 732 genes analyzed, 17 genes were found to be clinically meaningful. A score between 0 and 100 is generated based on these genes, with a higher score indicating more aggressive disease. The study found that a higher GPS score was independently predictive of a higher grade and stage on final surgical pathology, as well as metastasis and prostate cancer–specific survival, which are clinically meaningful outcomes, noted Dr. Yu.

The 31-gene CCP assay was tested on patients with low- to intermediate-risk disease from 3 centers, with a total of 582 patients. The study by Bishoff et al7 found that the CCP score could predict biochemical recurrence after prostatectomy. However, there is some controversy surrounding this assay, said Dr. Yu. A separate study found that although the CCP biomarker was associated with upgrading on final surgical pathology, it did not find an association with biochemical recurrence.8 Dr. Yu emphasized that biochemical recurrence is not necessarily an important outcome in prostate cancer, as it is not always associated with metastases or disease-specific survival.

Intermediate-Risk Prostate Cancer

For patients with intermediate-risk disease, Dr. Yu shared a post hoc analysis of an RTOG cohort that was conducted to validate the GC assay in an intermediate-risk population undergoing radiation therapy without hormone therapy. The study found that the GC assay was prognostic for various endpoints, including biochemical recurrence, distant metastasis, prostate cancer–specific survival, metastasis-free survival, and overall survival.9 Patients with a low GC score had a 10-year risk of distant metastasis of 4%, whereas those with a high GC score had a 16% risk. According to Dr. Yu, this suggests that patients with a low GC score may be able to avoid the toxicity of androgen-deprivation therapy (ADT).

High-Risk Prostate Cancer

For patients with high-risk prostate cancer undergoing radiation therapy, said Dr. Yu, a decision should be made between long-term and short-term ADT. A post hoc meta-analysis of 3 studies validated the Decipher test in patients with high-risk disease.10 The study found that Decipher was an independent prognosticator for clinically significant endpoints such as distant metastases, prostate cancer–specific mortality, and overall survival. “Interestingly, the absolute benefit of long-term compared with short-term ADT was approximately 11% in those with a high GC score and only 3% in those with a low GC score,” said Dr. Yu.

Although current NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Prostate Cancer still recommend long-term ADT for high-risk patients outside of clinical trials,11 shared decision-making can be considered for patients with a low GC score. Ongoing studies are investigating the potential for ADT de-escalation in some high-risk patients.

Another clinical scenario involves deciding whether a patient should receive adjuvant therapy after prostatectomy. The GC score was retrospectively validated in a postprostatectomy cohort.12 In a retrospective dataset of >300 patients who underwent radical prostatectomy without adjuvant therapy until development of metastasis, the risk of metastasis was 12% in those with a low GC score compared with 47% in those with a high GC score.

According to Dr. Yu, this information can be used to counsel patients, suggesting that adjuvant therapy may be considered in patients with a higher risk. Even in this high-risk cohort, however, Dr. Yu cautioned that Decipher is a prognosticator, not a predictor.

“It is unclear whether treating these patients earlier in the adjuvant setting compared with early salvage will result in clinically important long-term outcomes,” she stated. “Nonetheless, the test can provide additional information and may be useful when counseling patients.”

ArteraAI: MMAI

Finally, Dr. Yu introduced a new category of risk stratification tools known as MMAI assays, which have garnered significant interest in this field. MMAI assays use digital imaging of biopsy tissue specimens, which are then processed through a deep learning model to extract features that may not be apparent to pathologists. “By combining these extracted features with clinical variables such as Gleason score, clinical T stage, baseline PSA, and patient age, the goal is to create a better predictor of a patient’s long-term prognosis and determine if this risk stratifier is superior to NCCN risk groups,” said Dr. Yu.

A notable study that used data from 5 phase III RTOG studies found that the machine learning model was superior to the NCCN risk groups in discriminating important outcomes such as biochemical recurrence, metastasis, prostate cancer–specific mortality, and overall survival.13

“The AI model’s C-statistic (area under the curve [AUC]) was an impressive 0.83 compared with 0.72 for NCCN [risk groups],” said Dr. Yu. “However, the clinical utility of this technology is still unclear, as it is relatively new and in the pipeline.” Although specific cutoff scores to help direct treatment are not yet available, Dr. Yu suggested this is something that may be available in the future.

Another study investigated the use of artificial intelligence in a cohort of patients undergoing radiation therapy.14 Digital pathology images and clinical data from >5,000 patients enrolled in these studies were used to run through an MMAI agnostic model. Patients were labeled as model-positive or model-negative. In model-positive patients, adding ADT significantly reduced distant metastasis, whereas in model-negative patients, ADT did not provide any benefit.

“This exciting work may shape the future of clinical management for these patients,” Dr. Yu concluded.

References

  • 1.

    Klotz L, Vesprini D, Sethukavalan P, et al. Long-term follow-up of a large active surveillance cohort of patients with prostate cancer. J Clin Oncol 2015;33:272277.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 2.

    Reese AC, Pierorazio PM, Han M, et al. Contemporary evaluation of the National Comprehensive Cancer Network prostate cancer risk classification system. Urology 2012;80:10751079.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 3.

    Nguyen PL, Haddad Z, Ross AE, et al. Ability of a genomic classifier to predict metastasis and prostate cancer-specific mortality after radiation or surgery based on needle biopsy specimens. Eur Urol 2017;72:845852.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4.

    Spratt DE, Zhang J, Santiago-Jiménez M, et al. Development and validation of a novel integrated-genomic risk group classification for localized prostate cancer. J Clin Oncol 2018;36:581590.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 5.

    Vince RA Jr, Jiang R, Qi J, et al. Impact of Decipher Biopsy testing on clinical outcomes in localized prostate cancer in a prospective statewide collaborative. Prostate Cancer Prostatic Dis 2022;25:677683.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6.

    Klein EA, Cooperberg MR, Magi-Galluzzi C, et al. A 17-gene assay to predict prostate cancer aggressiveness in the context of Gleason grade heterogeneity, tumor multifocality, and biopsy undersampling. Eur Urol 2014;66:550560.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 7.

    Bishoff JT, Freedland SJ, Gerber L, et al. Prognostic utility of the cell cycle progression score generated from biopsy in men treated with prostatectomy. J Urol 2014;192:409414.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8.

    Cooperberg MR, Cowan JE, Lindquist KJ, et al. Multiple tissue biomarkers independently and additively predict prostate cancer pathology outcomes. Eur Urol 2021;79:141149.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9.

    Spratt DE, Liu VYT, Michalski J, et al. Genomic classifier performance in intermediate-risk prostate cancer: results from NRG Oncology/RTOG 0126 randomized phase 3 trial. Int J Radiat Oncol Biol Phys 2023;117:370377.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 10.

    Nguyen PL, Huang HR, Spratt DE, et al. Analysis of a biopsy-based genomic classifier in high-risk prostate cancer: meta-analysis of the NRG Oncology/Radiation Therapy Oncology Group 9202, 9413, and 9902 phase 3 randomized trials. Int J Radiat Oncol Biol Phys 2023;116:521529.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11.

    Schaeffer EM, Srinivas S, Adra N, et al. NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Prostate Cancer. Version 3.2024. Accessed April 2, 2024. To view the most recent version, visit https://www.nccn.org

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 12.

    Ross AE, Johnson MH, Yousefi K, et al. Tissue-based genomics augments post-prostatectomy risk stratification in a natural history cohort of intermediate- and high-risk men. Eur Urol 2016;69:157165.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 13.

    Esteva A, Feng J, van der Wal D, et al. Prostate cancer therapy personalization via multi-modal deep learning on randomized phae III clinical trials. NPJ Digit Med 2022;5:71.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 14.

    Spratt DE, Tang S, Sun Y, et al. Artificial intelligence predictive model for hormone therapy use in prostate cancer. NEJM Evid 2023;2:EVIDoa2300023.

Disclosures: Dr. Yu has disclosed serving as a consultant for AngioDynamics; and owning equity interest/stock options in Novo Nordisk.

Correspondence: Alice Yu, MD, MPH, Moffitt Cancer Center, 12902 Magnolia Drive, Tampa, FL 33612. Email: alice.yu@moffitt.org
  • Collapse
  • Expand
  • Figure 1.

    Localized prostate cancer: disease spectrum.

    Abbreviations: ADT, androgen deprivation therapy; RP, radical prostatectomy.

  • Figure 2.

    Clinical scenarios where molecular biomarkers can be helpful.

    Abbreviations: ADT, androgen deprivation therapy; CCP, cell-cycle progression; GC, genomic classifier; GPS, Genomic Prostate Score; LT-ADT, long-term androgen deprivation therapy; ST-ADT, short-term androgen deprivation therapy.

  • 1.

    Klotz L, Vesprini D, Sethukavalan P, et al. Long-term follow-up of a large active surveillance cohort of patients with prostate cancer. J Clin Oncol 2015;33:272277.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 2.

    Reese AC, Pierorazio PM, Han M, et al. Contemporary evaluation of the National Comprehensive Cancer Network prostate cancer risk classification system. Urology 2012;80:10751079.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 3.

    Nguyen PL, Haddad Z, Ross AE, et al. Ability of a genomic classifier to predict metastasis and prostate cancer-specific mortality after radiation or surgery based on needle biopsy specimens. Eur Urol 2017;72:845852.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4.

    Spratt DE, Zhang J, Santiago-Jiménez M, et al. Development and validation of a novel integrated-genomic risk group classification for localized prostate cancer. J Clin Oncol 2018;36:581590.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 5.

    Vince RA Jr, Jiang R, Qi J, et al. Impact of Decipher Biopsy testing on clinical outcomes in localized prostate cancer in a prospective statewide collaborative. Prostate Cancer Prostatic Dis 2022;25:677683.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6.

    Klein EA, Cooperberg MR, Magi-Galluzzi C, et al. A 17-gene assay to predict prostate cancer aggressiveness in the context of Gleason grade heterogeneity, tumor multifocality, and biopsy undersampling. Eur Urol 2014;66:550560.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 7.

    Bishoff JT, Freedland SJ, Gerber L, et al. Prognostic utility of the cell cycle progression score generated from biopsy in men treated with prostatectomy. J Urol 2014;192:409414.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8.

    Cooperberg MR, Cowan JE, Lindquist KJ, et al. Multiple tissue biomarkers independently and additively predict prostate cancer pathology outcomes. Eur Urol 2021;79:141149.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9.

    Spratt DE, Liu VYT, Michalski J, et al. Genomic classifier performance in intermediate-risk prostate cancer: results from NRG Oncology/RTOG 0126 randomized phase 3 trial. Int J Radiat Oncol Biol Phys 2023;117:370377.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 10.

    Nguyen PL, Huang HR, Spratt DE, et al. Analysis of a biopsy-based genomic classifier in high-risk prostate cancer: meta-analysis of the NRG Oncology/Radiation Therapy Oncology Group 9202, 9413, and 9902 phase 3 randomized trials. Int J Radiat Oncol Biol Phys 2023;116:521529.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11.

    Schaeffer EM, Srinivas S, Adra N, et al. NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Prostate Cancer. Version 3.2024. Accessed April 2, 2024. To view the most recent version, visit https://www.nccn.org

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 12.

    Ross AE, Johnson MH, Yousefi K, et al. Tissue-based genomics augments post-prostatectomy risk stratification in a natural history cohort of intermediate- and high-risk men. Eur Urol 2016;69:157165.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 13.

    Esteva A, Feng J, van der Wal D, et al. Prostate cancer therapy personalization via multi-modal deep learning on randomized phae III clinical trials. NPJ Digit Med 2022;5:71.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 14.

    Spratt DE, Tang S, Sun Y, et al. Artificial intelligence predictive model for hormone therapy use in prostate cancer. NEJM Evid 2023;2:EVIDoa2300023.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 471 471 471
PDF Downloads 315 315 315
EPUB Downloads 0 0 0