Causal Inference in Medical Device Trials: A Case Study

• Activity Number: 118
• Type: Topic Contributed
• Date/Time: Monday, August 4, 2003 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Heng Li*+
• Companies: FDA/CDRH
• Keywords: causal inference ; medical device ; clinical trial

Abstract

Randomization has been a cornerstone of scientific experiments, especially those involving humans. In practice the eventual implementation of a randomized experiment may deviate from its original plan for various reasons. A well-known kind of such deviations is the so-called noncompliance of enrolled patients in clinical trials. Another, perhaps less well-known, example in clinical trials involves patient enrollment. In the medical devices arena, sometimes it may be difficult to enroll patients from the population identified in the original protocol into a trial containing a no-treatment arm after the availability of a device having similar indications as the investigational device. When this happens, another patient cohort may be created in which subjects are randomized to treatments with the investigational device and the device that has just become available on the market. A study planned as a randomized controlled trial comparing two treatments thus becomes two randomized trials on two possibly very different patient populations involving three treatments. We discuss how, in such an event, causal inference may be made on the safety and efficacy of the investigational device.