participants in the trial can be justified, the presence of a robust safety-monitoring plan and other mechanisms for minimizing risk and harm to participants, and the equitable selection of participants.

Determining whether the other conditions are satisfied entails a thorough assessment of the evidence available and potentially available from observational studies. That assessment would be expected to address a number of questions, such as the following:

•  What are the limitations of the available evidence and the studies that provided it, and how important are these limitations?

•  Have possible uses of all existing data pertaining to the public health question been adequately explored?

•  What new information could FDA reasonably expect to obtain from new observational studies?

•  Would unacceptably large knowledge gaps or levels of uncertainty remain if FDA relied on observational studies or required additional observational studies to be conducted?

That determination also requires assessment of the likelihood that an RCT can provide the missing information with such questions as these:

•  Is it likely that a sufficient number of the eligible research participants could be recruited?

•  Is the expected timeframe of such a study to return information sufficient to inform FDA’s policy decision?

•  Would an appropriately designed RCT entail serious practicability issues?

DESIGN, ANALYTIC, AND ETHICAL CONSIDERATIONS IN SELECTING SPECIFIC OBSERVATIONAL AND RCT DESIGNS TO REQUIRE

Observational Studies

The determination by FDA that the additional evidence it needs to resolve a public health question can be obtained with observational studies presupposes that FDA has already identified a specific type of observational design that is both capable of generating the needed data and can be feasibly and ethically implemented. The success of an observational study in generating valid and reliable evidence requires access to sufficient high-quality data on the interventions compared, the outcomes of interest, and potential confounders to allow researchers to use analytic techniques that can approximate randomization to intervention assignment and preserve the study’s implications for causal inference.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement