Concordance Study in Hepatectomy Recommendations Between Watson for Oncology and Clinical Practice for Patients with Hepatocellular Carcinoma in China

Background With the improvement in diagnostic imaging, perioperative care and surgical technique, the indications and complexity of liver resections have developed. However, the surgical indications remain controversial especially for some complex or advanced hepatocellular carcinomas. This study was designed to evaluate the concordance between hepatectomy recommendations proposed by Watson for Oncology, a cognitive technology providing decision support, and those determined by surgeons in our center for patients with hepatocellular carcinoma. Methods We retrospectively reviewed 243 patients with hepatocellular carcinoma who were recommended for surgical treatment and received hepatectomy between 2008 and 2016 at the Zhujiang Hospital of Southern Medical University. Watson for Oncology classified the treatment options into three categories: recommended, for consideration and not recommended. Treatment recommendations were considered concordant if the hepatectomy recommendations were designated “recommended” or “for consideration” by Watson for Oncology. The factors potentially affecting concordance rate were also analyzed in our study. Results The hepatectomy recommendations of 174 patients were concordant. There were significant differences in the coincidence rate between concordant group and discordant group considering tumor numbers ( P = 0.006), extension of hepatectomy ( P = 0.009) and BCLC staging system ( P < 0.001). Lower degrees of concordance were observed in patients with multiple tumors, major hepatectomy and portal hypertension by using logistic regression analysis (OR = 0.309, P = 0.004; OR = 0.384, P = 0.004; and OR = 0.376, P = 0.022, respectively). Conclusion The concordance between Watson for Oncology and surgeons’ hepatectomy recommendation for hepatocellular carcinoma was only 72%. Differences in practice patterns for HCC between the USA (where Watson for Oncology was calibrated) and China may be the major cause of discordance. Watson for Oncology still requires further improvement and localization to be widely applied in China.