Background: To the authors' knowledge, the indications for radioactive iodine (RAI) therapy in patients with differentiated thyroid carcinoma (DTC) are unclear; treatment decisions are based on physician judgment. The objective of the current study was to identify the degree of concordance between postsurgical RAI therapy recommended by Watson for Oncology (WFO), a clinical decision support system for oncological therapy, and that recommended by physicians for patients with DTC.
Methods: The current retrospective cohort study included 207 patients with DTC who underwent thyroidectomy between 2017 and 2018. Treatment recommendations were considered concordant if WFO rendered recommendations consistent with those of the physicians.
Results: Treatment recommendations were concordant for 160 patients (77%). The concordance rate significantly differed according to the American Thyroid Association (ATA) risk category (P < .001) and American Joint Committee on Cancer TNM stage (seventh edition; P = .004). Logistic regression analysis demonstrated that treatment recommendations were significantly less likely to be concordant in patients with ATA intermediate-risk and stage III disease compared with those with ATA low-risk and stage I disease (odds ratio, 0.16 [P < .001] and OR, 0.35 [P = .004], respectively).
Conclusions: The authors believe the concordance rate between postsurgical RAI therapy recommendations rendered by WFO and those rendered by physicians was too low to justify adopting WFO for the comprehensive screening of patients with DTC. This is particularly true among patients with ATA intermediate-risk and stage III disease, reflecting differences in practice patterns between the United States (where WFO was calibrated) and Korea. Hence, WFO is not a substitute for physicians, and also may require regional customization to improve its assistive capability.
Keywords: Watson for Oncology; artificial intelligence; iodine radioisotopes; thyroid carcinoma.
© 2019 American Cancer Society.