Determining the risk level of localized prostate cancer (PCa) at the time of diagnosis is often closer to an art than a science. Is it low risk, favorable intermediate risk, unfavorable intermediate risk, or high risk? Precision matters, because it determines the treatment plan’s projected probability of success. If the treatment isn’t aggressive enough, there’s a higher chance of recurrence. On the other hand, if it’s overkill, there’s a higher chance of side effects that interfere with quality of life.

For PCa still contained in the gland, today’s standard risk group criteria as spelled out by the National Comprehensive Cancer Network (NCCN) are as follows:

• low risk: T1–T2a, Gleason score ≤6, and PSA <10 ng ml
• intermediate risk: T2b–T2c or Gleason score 7 or PSA 10–20 ng ml
• high risk: T3a or Gleason score 8–10 or PSA >20 ng ml^[i]

Since the Gleason score is assigned by a pathologist who evaluates the biopsy tissues by microscopic examination, there’s room for human error (a major reason why patients are advised by PCa advocacy groups to always get a second opinion on their biopsy slides). Is there a more reliable and consistent way to stratify an individual’s risk level based on his biopsy specimens and other clinical factors?

Clinical trial tests Artificial Intelligence (AI) model

That question was put to the test by a research team from several academic centers and an AI development company, Altera. Their study was designed to assess the ability of a multimodal artificial intelligence (MMAI) model to outperform the NCCN risk group criteria.^[ii] Their population consisted of 9,787 PCa patients whose localized disease had been treated with radiation, hormone therapy or chemotherapy in previous clinical trials. All patient cases had the original diagnostic digital biopsy slide images and clinical data on record, and had then been followed for an average of 7.9 years. Based on the original diagnostic information, the MMAI model was trained to classify risk level and calculate the probability of distant metastasis at 10 years, and the team then compared those results against the projections of the NCCN classification system.

Here’s what the authors wrote:

The median follow-up for censored patients was 7.9 years. According to NCCN risk categories, 30.4% of patients were low-risk, 25.5% intermediate-risk, and 44.1% high-risk. The MMAI risk classification identified 43.5% of patients as low-risk, 34.6% as intermediate-risk, and 21.8% as high-risk. MMAI reclassified 1,039 (42.0%) patients initially categorized by NCCN.

The risks of 10-year metastasis were comparable in both systems for the low-risk groups (1.7% for NCCN and 3.2% for MMAI), even though the MMAI low-risk group was larger. While the NCCN high-risk group had an overall predicted 10-year metastasis rate of 16.6%, the MMAI system had further reclassified some of these patients as low or intermediate risk, thus breaking down the 10-year metastasis projection into 3.4% (low risk), 8.2% (intermediate risk) and 26.3% (high risk).

Conclusion

The researchers concluded, “The MMAI risk grouping system expands the population of men identified as having low metastatic risk and accurately pinpoints a high-risk subset with elevated metastasis rates.” Along with greater accuracy, AI also brings consistency in identifying biopsy characteristics—unlike human judgment. This is good news for recently diagnosed patients, since accurate stratification helps avoid overtreatment of low-risk disease, and undertreatment of higher risk disease. As the authors point out, the shared decision-making process between doctor and patient benefits from better knowledge about the future behavior of each man’s prostate cancer.

Although the MMAI model is still at the clinical testing stage, the fact that this was a Phase III trial means it’s close to gaining FDA approval and therefore clinical availability.

NOTE: This content is solely for purposes of information and does not substitute for diagnostic or medical advice. Talk to your doctor if you are experiencing pelvic pain or have any other health concerns or questions of a personal medical nature.

References

[i] Xu H, Zhu Y, Dai, B, Ye DW. National Comprehensive Cancer Network (NCCN) risk classification in predicting biochemical recurrence after radical prostatectomy: a retrospective cohort study in Chinese prostate cancer patients. Asian Journal of Andrology 20(6):p 551-554, Nov–Dec 2018.
[ii] Tward JD, Huang HC, Esteva A et al. Prostate Cancer Risk Stratification in NRG Oncology Phase III Randomized Trials Using Multimodal Deep Learning With Digital Histopathology. JCO Precis Oncol 8, e2400145(2024).