臨床試驗中的知情同意: 理論、政策與研究

Abstract

本論文從資訊行為與資訊設計的視角，檢視1995年至2024年間臨床試驗「知情同意」的演變。儘管法規長期要求充分揭露，受試者對隨機分配、安慰劑對照，以及研究與臨床照護之區別等核心概念的理解仍普遍不足。這一持續存在的落差顯示，知情同意並非單純的合規問題，而是根植於資訊設計與人類行為的結構性挑戰。 研究採系統性的多元方法，整合三個層面：（一）文獻回顧以梳理三十年的理論演進；（二）比較分析美國與歐盟之法規架構；（三）實證的領域分析，結合潛在狄利克雷分配主題模型與共詞網絡，檢視5,978篇學術出版品。結果顯示此領域正處於「複雜且分散的適應」：出版量自2004年的95篇增至2024年的480篇，成長約五倍；研究社群由13個獨立群組擴增為26個。以科技為核心的子領域，如機器人、人工智慧與數位健康，在2015年後才作為獨立主題浮現，實證支持理論與監管正朝向以設計為基礎的解方轉型。 本論文提出一套整合倫理理論、法規科學與資訊設計的概念框架，將知情同意定位為一個動態系統：它需要以參與者為中心的設計、科技創新與持續調適，而非僅止於靜態揭露。分析亦顯示，美國與歐盟的做法正收斂於強調透明度、風險比例性與設計品質，超越單一的合規導向。 本研究為研究者、監管機關與實務工作者提供可操作的建議，以因應日益複雜的試驗設計、擴張中的大數據資料庫，以及人工智慧在臨床研究中的應用。

This dissertation focuses on informed consent in clinical trials as a dimension of information behavior and design and its altogether development from 1995 to 2024. Even when consent procedures follow regulations, participants still struggle to understand randomization, placebos, and the difference between clinical care and research care. This lack of understanding about informed consent points toward several pervasive issues of structural information design and the human behavior intertwined with it, instead of being just a legal matter. This dissertation follows an integrated approach that applies multiple methods: conducting a literature review on three decades of evolution of the field, a comparative analysis of the U.S. and European Union regulatory systems, and an empirical domain analysis that combines Latent Dirichlet Allocation (LDA) topic modeling and co-word network analysis of 5,978 research papers. The results outline a domain that is in a stage of “complex and fragmented adaptation.” The output of publications increased fivefold (from 95 in 2004 to 480 in 2024) and the number of unique research communities doubled from 13 to 26. Domains focused on technology and including robotics, artificial intelligence, and advanced digital health altogether began to emerge as distinct research domains only after 2015, providing empirical evidence to the theoretical and regulatory changes emphasizing design-centered solutions. This dissertation’s main impact is a refined conceptual framework that integrates ethical theory, information design, and regulatory dimensions into a unified model. Such a refined framework positions informed consent not as a static disclosure, but as a dynamic system that needs continuous adaptation, participant-centered design, and technological innovation. This dissertation’s analyses also show a convergence of the U.S. and E.U. approaches as both are shifting towards a greater focus on transparency, risk-proportionate related regulations, and design quality over mere compliance. With these in mind, such regulatory, research, and practitioner-based insights are potentially altogether effective in clinical research that involves complex trials, expansive data sets, and artificial intelligence applications.