Targeting the Target Trial; a review of target trial element operationalization

Abstract

Researchers often make use of Randomized Controlled Trails (RCT) to uncover whether a treatment truly works or not. By randomly assigning patients to either control (patient does not receive treatment of interest) or treatment (patient does receive treatment of interest) researchers can estimate how effective the treatment is. These trials, however are generally extremely expensive, can take years and due to ethical or practical reasons, are impossible to carry out. In recent years a new method has emerged: The Target Trial Emulation (TTE) framework. Instead of running trials, researchers make use of already existing data, such as hospital records, insurance claims, recorded prescriptions to determine which patient was assigned the treatment arm and who to the control arm, and then estimate what the difference in outcomes would have been.

This project reviewed 100, recently published studies using the TTE framework to understand how researchers “built” their emulation from available data. For example, how did the researchers in each study identify the people that were included and excluded from the trial? and how did they identify the patients in the control and treatment arm? and how did the researchers look for the outcomes? This project examined how researchers identified these key aspects from the data that they used.

This review found that while most studies use highly detailed hospital, electronic health record, or intensive care unit data, quite a few studies do not provide a method by which their target population was identified, making it extremely difficult for other researchers to replicate their research. Even though many studies relied on highly detailed data, studies using data with less detail can also produce reliable results when the research question aligns with the data.

Overall, this study shows that it is important for researchers to provide clarity and transparency in reporting how patients, treatments, and outcomes are identified from the data source. Making future research reliable, credible, trustworthy and consistent.