Concordance study between IBM Watson for Oncology (WFO) and clinical practice for breast and lung cancer patients in China

Abstract

Background: IBM Watson for Oncology (WFO)is aMemorial Sloan Kettering-trained cognitive computing system that uses natural language processing (NLP) to ingest data in structured and unstructured formats. WFO provides physicians with evidence-based treatment options and ranks them into three categories for oncology treatment decision support. This study examined the degree of concordance between treatment recommendation supposed byWFO and amulti-disciplinary tumor board in The Affiliated Hospital of Qingdao University. Methods: Retrospective data from 152 cancer patients were ingested into WFO from Aug, 2015 to Jul, 2017. WFO recommendations were provided in 3 categories: “Recommended”, “For Consideration”, and “Not Recommended.” Concordance was analyzed by comparing the treatment decisions proposed by WFO to those of the multi-disciplinary tumor board. Concordance was achieved when the multi-disciplinary tumor boards' treatment decisions were in the “Recommended” or “For Consideration” categories in WFO. Results: A total of 152 cancer cases were assessed, 119 were breast cancers and 33 were lung cancers. Overall, treatment recommendations were concordant in 79.0% (94 of 119) of breast cancers and 96.9% (32 of 33) in lung cancers. Further analysis the concordance difference between breast and lung cancer, the main reason for this situation is due to the payment capacity in targetted therapy. Some Her2-positive breast cancer patients cannot afford trastuzumab therapy which is regarded as standard in guidelines and WFO. The concordance is 90.3% in Her2-negative patients and 61.7%in Her2-positive patients. Conclusions: Treatment recommendations made byWFO were highly concordant with clinical practice by the affiliated hospital of Qingdao University in multiple types of carcinoma. This cognitive system is very useful in decision making support and multi-disciplinary tumor board.