Artificial Intelligence And Clinician Decision Making For Postoperative Adjuvant Therapy For Breast Cancer Patients In Consistency With CSCO Guidelines

Objectives:To explore the consistency of the Chinese artificial intelligence system(Haixin Zhihui)and clinicians’ decision making on postoperative adjuvant treatment for breast cancer patients with the Chinese Society of Clinical Oncology’s breast cancer treatment guidelines,to analyze the factors affecting the consistency,to find the reasons for the inconsistency with the guidelines,and to study the role of artificial intelligence systems in clinical practice as an aid to physicians’ decision making.Methods:A total of 120 post-operative breast cancer patients from January to December2022 will be collected and their basic information such as age,menstrual status and gender,as well as the main immunohistochemical indexes and lymph node metastasis of post-operative breast cancer tissues will be entered into the Haixin Zhihui AI system to apply for the AI diagnosis and treatment report.The treatment plan will show "recommended" and "not recommended",using the 2022 CSCO breast cancer treatment guidelines as reference.The treatment plan was determined to be consistent with the guideline when the system recommendation was consistent with the guideline level I recommendation or the AI system and the guideline did not recommend relevant treatment plan.At the same time,the clinician’s clinical plan was compared with the guidelines,and the clinician’s treatment plan was judged to be consistent with the guidelines when the clinician’s treatment plan was the same as the "Level I recommendation" of the CSCO guidelines or when neither the clinician nor the guidelines recommended the relevant treatment plan,while the rest of the cases were judged to be non-compliant.Results:1.Consistency of clinicians’ decisions with guidelines: 93.3% and 92.5% of clinicians’ decisions for postoperative radiotherapy and endocrine for breast cancer patients were consistent with guidelines,respectively;83.3% of targeted therapy was consistent with guidelines;and 83% of chemotherapy was consistent with guidelines.A univariate subgroup analysis showed that clinician decision making for postoperative chemotherapy versus guideline concordance was influenced by patient age and histologic grading.Multiple regression analysis showed that age was an independent risk factor influencing clinicians’ decision making about consistency of postoperative chemotherapy with guidelines in breast cancer patients.2.Consistency of CSCO AI decisions with guidelines: CSCO AI decisions for breast cancer patients with postoperative radiotherapy and endocrine regimens were both 100% consistent with guidelines;targeted therapy was 94.4% consistent with guidelines;and chemotherapy was 92.5% consistent with guidelines.A univariate subgroup analysis showed that CSCO AI decision chemotherapy-guideline concordance was influenced by histologic grade as well as menstrual status.Multiple regression analysis showed that lymph node status and type of pathology were independent risk factors affecting the consistency of chemotherapy regimens with guidelines in CSCO AI decision making.3.CSCO AI decision vs.clinician agreement: CSCO AI decision postoperative chemotherapy regimens for breast cancer patients were more consistent with guidelines than clinicians(92.5% vs.83%),and the difference was statistically significant.CSCO AI decision targeted therapy was more consistent with guidelines than clinicians(94.4% vs.83.3%),and the difference between the two was not statistically significant.CSCO AI decision postoperative radiotherapy and endocrine regimens were both more consistent with guidelines than clinicians(100% vs.93.3%;100% vs.92.5%).Conclusions:1.The CSCO AI decision postoperative treatment plan for breast cancer patients had a high level of agreement with the guidelines.Among them,the decision-making radiotherapy and endocrine therapy regimens were 100% consistent with the guidelines.In different subgroups,the consistency of CSCO AI decision making for postoperative chemotherapy and targeted therapy was higher than that of clinicians.To some extent,the study suggests that CSCO AI can be useful for clinicians in making decisions about breast cancer treatment options.2.The development of genetic testing programs,clinicians’ personal experience and customary medication regimens that differ from the guidelines,patients’ personal treatment wishes,and their ability to pay may lead to inconsistencies between clinicians’ post-operative adjuvant treatment regimens for breast cancer patients and the guidelines,etc.The interaction between CSCO AI and physicians can make clinical treatment decisions more accurate and appropriate for patients.3.Although CSCO AI has unparalleled advantages,there are still shortcomings in its actual clinical use,such as the logical priority of AI calculation,omission or error of case information input by operators,and inability to consider the actual situation of patients in batch processing,which may lead to inconsistency between CSCO AI decision making and guidelines.More data studies are needed to verify the consistency of CSCO AI with guidelines and its impact on clinicians’ decision making in order to explore the best clinical practice model.