An artificial intelligence (AI)-derived model can identify prostate cancer patients who are likely to benefit from short-term androgen deprivation therapy (ADT), according to research published in NEJM Evidence.
To develop the model, researchers used digitized pretreatment prostate cancer tissue pathology slides and clinical data from 5727 patients who were treated with radiotherapy, with or without ADT, in 5 phase 3 trials.
The researchers validated the AI-derived model using data from the NRG Oncology/RTOG 9408 trial. This trial included 1594 patients with prostate cancer, 56% of whom had intermediate-risk disease. All patients were randomly assigned to receive radiotherapy with short-term ADT (n=788) or radiotherapy alone (n=806).
When the researchers applied the AI-derived model, 543 patients were identified as model-positive. These patients had a significantly lower 15-year distant metastasis rate with ADT than with radiotherapy alone — 4.0% and 14.4%, respectively (subdistribution hazard ratio [sHR], 0.34; 95% CI, 0.19-0.63; P <.001).
For patients who were identified as model-negative, there was no significant difference in 15-year distant metastasis between patients who received ADT and those who did not. The 15-year distant metastasis rate was 6.9% with ADT and 7.4% with radiotherapy alone (sHR, 0.92; 95% CI, 0.59-1.43; P =.71).
Among patients who were model-positive, there was a reduction in 15-year prostate cancer-specific mortality with ADT. It was 2.6% with ADT and 12.7% with radiotherapy alone (sHR, 0.28; 95% CI, 0.14-0.57).
Among patients who were model-negative, however, the 15-year prostate cancer-specific mortality was similar between the treatment groups. It was 5.3% with ADT and 6.5% with radiotherapy alone (sHR, 0.74; 95% CI, 0.45-1.22).
Based on these results, the researchers concluded that their AI-based model was able to identify prostate cancer patients who are likely to benefit from short-term ADT.
Disclosures: This research was partly supported by Artera, Inc. Some study authors declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of disclosures.
Spratt DE, Tang S, Sun Y, et al. Artificial intelligence predictive model for hormone therapy use in prostate cancer. NEJM Evidence. Published online June 29, 2023. doi:10.1056/EVIDoa2300023