Department of Statistics and Actuarial Science, 25809The University of Hong Kong, Hong Kong, China.
Stat Methods Med Res. 2023 Feb;32(2):229-241. doi: 10.1177/09622802221133555.
Randomized controlled trials (RCTs) have been widely recognized as the gold standard to infer the treatment effect in clinical research. Recently, there has been growing interest in enhancing and complementing the result in an RCT by integrating real-world evidence from observational studies. The unit information prior (UIP) is a newly proposed technique that can effectively borrow information from multiple historical datasets. We extend this generic approach to synthesize the non-randomized evidence into a current RCT. Not only does the UIP only require summary statistics published from observational studies for ease of implementation, but it also has clear interpretations and can alleviate the potential bias in the real-world evidence via weighting schemes. Extensive numerical experiments show that the UIP can improve the statistical efficiency in estimating the treatment effect for various types of outcome variables. The practical potential of our UIP approach is further illustrated with a real trial of hydroxychloroquine for treating COVID-19 patients.
随机对照试验（RCT）已被广泛认为是推断临床研究中治疗效果的金标准。最近，人们越来越感兴趣的是通过整合来自观察性研究的真实世界证据来增强和补充 RCT 的结果。单位信息先验（UIP）是一种新提出的技术，它可以有效地从多个历史数据集借用信息。我们将这种通用方法扩展到将非随机证据综合到当前的 RCT 中。UIP 不仅只需要来自观察性研究的汇总统计信息，便于实施，而且具有明确的解释，并且可以通过加权方案减轻真实世界证据中的潜在偏差。广泛的数值实验表明，UIP 可以提高各种类型的结果变量估计治疗效果的统计效率。我们的 UIP 方法的实际潜力通过羟氯喹治疗 COVID-19 患者的真实试验进一步说明了。
