Using Artificial Intelligence (Watson for Oncology) for Treatment Recommendations Amongst Chinese Patients with Lung Cancer: Feasibility Study.

BACKGROUND: Artificial intelligence (AI) is developing quickly in the medical field and can benefit both medical staff and patients. The clinical decision support system Watson for Oncology (WFO) is an outstanding representative AI in the medical field, and it can provide to cancer patients prompt treatment recommendations comparable with ones made by expert oncologists. WFO is increasingly being used in China, but limited reports on whether WFO is suitable for Chinese patients, especially patients with lung cancer, exist. Here, we report a retrospective study based on the consistency between the lung cancer treatment recommendations made for the same patient by WFO and by the multidisciplinary team at our center.

OBJECTIVE: The aim of this study was to explore the feasibility of using WFO for lung cancer cases in China and to ascertain ways to make WFO more suitable for Chinese patients with lung cancer.

METHODS: We selected all lung cancer patients who were hospitalized and received antitumor treatment for the first time at the Second Xiangya Hospital Cancer Center from September to December 2017 (N=182). WFO made treatment recommendations for all supported cases (n=149). If the actual therapeutic regimen (administered by our multidisciplinary team) was recommended or for consideration according to WFO, we defined the recommendations as consistent; if the actual therapeutic regimen was not recommended by WFO or if WFO did not provide the same treatment option, we defined the recommendations as inconsistent. Blinded second round reviews were performed by our multidisciplinary team to reassess the incongruent cases.

RESULTS: WFO did not support 18.1% (33/182) of recommendations among all cases. Of the 149 supported cases, 65.8% (98/149) received recommendations that were consistent with the recommendations of our team. Logistic regression analysis showed that pathological type and staging had significant effects on consistency (P=.004, odds ratio [OR] 0.09, 95% CI 0.02-0.45 and P<.001, OR 9.5, 95% CI 3.4-26.1, respectively). Age, gender, and presence of epidermal growth factor receptor gene mutations had no effect on consistency. In 82% (42/51) of the inconsistent cases, our team administered two China-specific treatments, which were different from the recommendations made by WFO but led to excellent outcomes.

CONCLUSIONS: In China, most of the treatment recommendations of WFO are consistent with the recommendations of the expert group, although a relatively high proportion of cases are still not supported by WFO. Therefore, WFO cannot currently replace oncologists. WFO can improve the efficiency of clinical work by providing assistance to doctors, but it needs to learn the regional characteristics of patients to improve its assistive ability.