(201) Specifying the target trial reveals exceptions to commonly accepted rules of target trial emulation

Background: The target trial emulation framework is a vital approach for designing valid and transparent real-world (RW) studies of the effectiveness and safety of medical treatments. In applying this framework, we have found instances where commonly accepted “rules” about randomization and valid RW study design require more nuanced consideration.

Objectives: To demonstrate how the target trial emulation framework can identify sources of confounding that would be uncontrolled even in the randomized target trial and understand when post-baseline information can be used to determine eligibility in RW studies.

Methods: We considered a hypothetical example in which we are interested in studying the RW effectiveness of the hepatitis B vaccine for preventing chronic hepatitis B. We created a target trial for emulation where individuals would be screened for acute and chronic infection at baseline, randomized to receive the hepatitis B vaccine or a placebo vaccine, and followed for chronic hepatitis B infection.

Results: Acquisition of an acute hepatitis B infection is a necessary precursor of chronic infection. However, even in the target trial, acute hepatitis B infection is not random or assignable. As a result, additional confounding control is required in both the hypothetical target trial and observational study. In addition, we can stipulate that we would screen all potential participants for acute infection at baseline in the target trial, but this information is unlikely to be available for all patients at the time of vaccination in RW data. As a result, we may need to use information on infections observed after vaccination to determine eligibility at baseline and ensure that patients are at risk of the outcome at the beginning of follow-up. While researchers are often cautioned against using post-baseline information to determine eligibility, the hypothetical target trial framework illustrates why this is permissible under these circumstances and does not lead to immortal time bias.

Conclusions: The target trial emulation framework is a valuable tool for RW study design and reveals exceptions to commonly-accepted rules for avoiding bias.