Using this registry along with a machine learning prediction model and web development platform, a web-based tool was created that they named ‘askMUSIC’. Predictor variables included: age, weight, co-morbidities, PSA level, Gleason score, number positive biopsy cores, and total biopsy cores sampled. The model, which identifies similar patients in the registry, predicts the likelihood of receiving prostatectomy, radiotherapy, surveillance, androgen deprivation, and watchful waiting. It relies on registry data from 2011-2015 and was validated using multinomial AUC and calibration curves.
Using this data, of the 11,456 men included, 44.7% underwent prostatectomy, 22.0% surveillance, 19.5% radiation, 8.8% androgen deprivation therapy, and 3.6% watchful waiting.
This patient tool is useful in providing insight into treatment choices to support informed decision making. This method will be beneficial in the future in determining both functional and oncologic outcomes to drive patient care.
Presented By: Gregory Auffenberg, MD, MS
Authors: Gregory Auffenberg, Shreyas Ramani, Khurshid Ghani, Brian Denton, Craig Rogers, Benjamin Stockton, David Miller, Karandeep Singh
Written By: David B. Cahn, DO, MBS, Fox Chase Cancer Center
at the 2017 AUA Annual Meeting - May 12 - 16, 2017 – Boston, Massachusetts, USA