人工智能(artificial intelligence,AI)已成为引领未来的战略性技术,也是中国未来发展的关键引擎。在医疗器械的创新研发中,AI已经在智能辅助诊断、智能辅助治疗、智能监护与生命支持等方面提供了关键支持,机器学习赋能设备软件功能(machine learning-enabled device software functions,ML-DSFs)已成为许多医疗器械的重要组成部分。近期,美国食品药品监督管理局(Food and Drug Administration,FDA)发布了《针对人工智能/机器学习赋能设备软件功能的预设变更控制计划上市提交建议的指南草案》,希望提供一个前瞻性方法来促进机器学习医疗器械的发展,在保证设备的持续安全性和有效性前提下,支持ML-DSF通过修改来迭代更新。该指南代表了最新的监管方向,特别有助于提升AI产品临床试验质量与效率,因此撰写本文加以详细介绍和解读,以利于借鉴国际先进监管理念和经验,促进产业健康发展和国际影响力提升。
本研究旨在探讨申办者在医疗器械临床试验中构建文件质量控制体系的必要性,并介绍基于量化评价的全面质量控制体系。该体系以国内相关规范为指导,涵盖研究中心和申办者文件夹的质控,通过评分机制客观评估文件质量,并结合质控员–监查员–研究者反馈环路以及质控员–质控体系负责人–质控体系环路,实现问题发现、整改和持续改进。超过85%的临床试验质控评分呈现出逐次上升的趋势,质控中问题的发生率也呈现出下降趋势。该体系有效提升了临床试验文件质量,为申办者提供了标准化、可操作的质控流程,为提升我国医疗器械临床试验整体管理水平提供了参考。This study aims to investigate the necessity for sponsors to establish a quality control system for clinical trial documentation in medical device clinical trials and to introduce a quantitatively assessed total quality control system. Guided by domestic regulations, this system encompasses the quality control of both investigator and sponsor Trial Master File (TMF). TMF quality can be assessed objectively through a scoring mechanism, which is integrated with a feedback loop involving quality controllers, monitors, and investigators, as well as a loop between quality controllers, quality control management, and the quality control system, to facilitate problem identification, rectification, and continuous improvement. Over 85% of clinical trial quality control scores showed a trend of successive increase, and the incidence of issues identified during quality control also demonstrated a decreasing trend. The system effectively enhances the quality of clinical trial documentation, providing sponsors with a standardized and operational quality control process, and serves as a reference for improving the overall management level of medical device clinical trials in our country.
