When based on extensive data sets composed of fully defined characteristics of the patients likely to utilize the model, AI can make an accurate prediction of outcome for an individual patient. These requirements were met by Spratt et al. in developing the first iteration of “My Prostate Test” by ArteraAI, in their article “Artificial Intelligence Predictive Model for Hormone Therapy in Prostate Cancer,” NEJM Evidence, June 2023.
Their model is a clinically available AI-enabled test (My Prostate Test) to predict the relative benefit of adding short-term hormone suppression to radiation in the primary treatment prostate cancer. To accomplish this, they integrated the digitized pre-treatment histopathological data, clinical characteristics, and long-term (8 yr) outcome data of 2024 patients with localized cancer from 5 Phase III prospective clinical trials comparing radiation therapy +/- 4 months adjuvant Lupron + an antiandrogen. In the study 87% had NCCN intermediate-risk cancer (Gleason Score 3+4). Their goal was to create a multimodal predictive model to determine the differential relative benefit of combining 4 months of adjuvant ADT with radiotherapy compared to radiotherapy alone in prolonging the time to developing distant metastases.
How does the algorithm work? : An illustration Based on Intermediate-Risk Patients
After receiving the digitalized histopathology data from the pre-treatment biopsy slides, ArteraAI applies their algorithm, analyzes the specimen, and evaluates the result with their large database, the repository of outcome records of the 5442 patients in the study, half receiving radiotherapy alone or in combination with ADT. Their algorithm creates a “predictive model” for each individual – either positive or negative.
Results: The men who received radiation (RT) and ADT who were predictive model positive benefited from ADT and experienced a 10-year distant metastasis (DM) estimate of 4% versus 14.4% for those men treated with radiation therapy only. In contrast in the model negative group there was no benefit from adding ADT to RT. Both the RT + ADT and the RT alone groups showed comparable 10-year DM rates, 6.9% and 7.4%.
Conclusion: The authors state that their “AI-based predictive model was able to identify prostate cancer patients, with predominantly intermediate-risk disease, who are likely to benefit from short-term ADT.” “Using this predictive model, we showed from the trial data that the majority of intermediate-risk patients did not benefit from ADT.”
Extending “My Prostate Test” to Address High-risk Patients:
Spratt et al., (Eur Urol 2024 Jul) extended the application of My Prostate Test to risk-stratify men with localized high-risk prostate cancer based on the analysis of digitalized histopathology slides of 426 patients from 3 randomized clinical trials. The study included men with clinical stage T-4, Gleason Score 8 – 10, PSA >20 ng/mL and primary Gleason pattern 5 with median follow-up of 10.4 years. The study endpoints were the development of distant metastases (DM) and Prostate Cancer Specific Mortality at 10 years.
Conclusion: Depending on a man’s individual characteristics the DM outcome raged from 8% for the lower 25% of men in the study to 26% for the highest quartile. As was also the case for men with intermediate-risk cancer, both analyses revealed the wide variation in response to treatment in these conventional NCCN risk groupings. “This scalable tool can provide physicians and patients with more personalized information for shared decision-making on treatment.”
The Latest Addition to “My Prostate Test”:
ArteraAI has expanded ”MyProstateTest” to include prediction for developing distant metastases 10 years after a prostatectomy and guidance for men considering active surveillance vs. active treatment. “The prognostic performance of the test has now been validated in a diverse cohort of patients, including those who have undergone active surveillance or radiation therapy or had a radical prostatectomy,” Trevor Royce MD ,MPH, medical director of ArteraAI. The new application of the test to active surveillance was based on a study by Esteva et al, “Prostate cancer therapy personalization via multi-modal deep learning on randomized phase III clinical trials.” Their artificial intelligence-based tool employed clinical data and digital histopathology of 5654 biopsies of men followed for a median of 11.4 years and predicts 5-and 10- year biochemical failure and distant metastases, identifying more aggressive disease that would warrant intervention. Their AI risk-stratifying tool was more accurate than the customary NCCN model. “Outfitted with digital scanners and internet access, any clinic could offer such capabilities, enabling global access to therapy personalization.”
To learn more about “My Prostate Test” and how to order a test, search “Advanced AI Prostate Cancer Test.”
BOTTOM LINE:
“My Prostate Test” (ArteraAI) provides individualized outcome predictions for prostatectomy, active surveillance and the efficacy of adjuvant hormone suppression supplementing primary radiation therapy to guide management decisions.