Early experience with Watson for Oncology: a clinical decision-support system for prostate cancer treatment recommendations

Abstract

Purpose Urological oncologists have difficulty providing optimal personalized care due to rapid alterations in scientific research results, medical advancements, and treatment guidelines. IBM’s Watson for Oncology (WFO) is an artificial intelligence clinical decision-support system that assists oncologists with evidence-based treatment recommendations. In the present study, we examined the level of concordance between the treatment recommendations for prostate cancer according to WFO and the actual treatments that the patients received in the department of urology. Methods We enrolled 201 patients who received prostate cancer treatment between January 2018 and June 2018. WFO provided treatment recommendations using clinical data in three categories: recommended, for consideration, and not recommended. These were compared with the actual treatments received by patients. Prostate cancer treatments were considered concordant if the received treatments were included in the “recommended” or “for consideration” categories by WFO. Results The patients’ mean age was 71.2 years. There were 60 (29.9%) and 114 (56.7%) patients with an Eastern Cooperative Oncology Group (ECOG) performance score ≥ 1 and non-organ confined disease (stage III/IV), respectively. The overall prostate cancer treatment concordance rate was 73.6% (“recommended”: 53.2%; “for consideration”: 20.4%). An ECOG performance score ≥ 1 and older age (≥ 75 years) were significantly associated with discordance (p = 0.001 and p = 0.026, respectively) on multivariate analysis. Conclusion In the present study, the treatment recommendations by WFO and the actual received treatments in the department of urology showed a relatively high concordance rate in prostate cancer patients.