Answers are grounded in this paper and cite their sources.
The primary goal of randomized controlled trials is typically to understand the impact of a treatment on an outcome(s) of interest. Randomization is a critical design component for making these types of cause-and-effect inferences, as it ensures balance between the treatment groups in terms of both known and unknown confounders (factors related to both exposure and outcome), which can introduce bias when imbalanced across exposure (treatment) groups. When comparing treatment groups in the primary analysis of a randomized controlled trial, participants are typically included with the group to which they were randomized, following a framework referred to as intention to treat (ITT). ITT intentionally does not account for postrandomization events (eg, protocol nonadherence) in order to maintain the integrity of the randomization. There are other non-ITT frameworks for randomized controlled trial analyses, including the common per-protocol, as-treated, and modified ITT analyses. This article summarizes ITT and non-ITT frameworks and the research questions which they address, discusses their trade-offs, and provides information about operationalizing randomized controlled trial analyses, including discussion of the available topical guidance, considerations for conducting analyses in the face of missing data, and evolving statistical methods for non-ITT analyses.
No takes yet. Share an insight, caveat, or question.
McCabe et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: