Some Considerations on Design and Analysis Plan on a Nonrandomized Comparative Study Using Propensity Score Methodology for Medical Device Premarket Evaluation

• Activity Number: 189 - SBR Showcase: Recent Advances in Statistical Methodology for Medical Product Development
• Type: Invited
• Date/Time: Tuesday, August 10, 2021 : 1:30 PM to 3:20 PM
• Sponsor: Statistics in Biopharmaceutical Research Journal
• Author(s): Nelson Lu* and Yunling Xu and Lilly Q. Yue
• Companies: FDA/CDRH and CDRH/FDA and FDA/CDRH
• Keywords: real-world data; propensity score; estimand; nonrandomized study; two-stage design

Abstract

Nonrandomized comparative studies have been widely proposed and conducted in the premarket safety and effectiveness evaluation of medical devices. A typical case is to compare results of a clinical study on an investigational device with those of control, which is formed based on prior clinical studies or real-world data. Propensity score methodology is often used to mitigate the bias caused by the imbalance of the patient baseline characteristics. Proper design and data analyses are essential in obtaining valid evidence. To maintain the objectivity of study design, a two-stage design needs to be implemented. To evaluate the data, analysis methods for analyses such as endpoint analysis, missing data handling, and subgroup analysis need to be carefully planned. In this talk, some considerations for the design and analysis plan will be discussed, and an example that mimic to cases that have been encountered will be provided to illustrate these considerations.