Clinical research & EBM · Reference
What is intention-to-treat analysis?
Intention-to-treat analysis is a principle for analysing a trial in which participants are kept in the group to which they were randomly assigned, regardless of whether they adhered to it. This preserves the benefit of randomisation and gives a realistic estimate of effectiveness.
Analyse as randomised
The core rule of intention-to-treat is "once randomised, always analysed". Every participant is counted in the group they were assigned to, even if they switched intervention, stopped early or did not follow the protocol. This matters because the protection that randomisation offers against confounding comes from the assignment itself. Removing or re-classifying participants after the fact — for example dropping those who did not adhere — can reintroduce systematic differences between groups, because non-adherence is often related to prognosis. ITT keeps the randomised comparison intact.
ITT versus per-protocol
The main contrast is with per-protocol analysis, which includes only participants who adhered closely to their assigned intervention. Per-protocol analysis estimates the effect under ideal adherence (efficacy), but by selecting on post-randomisation behaviour it risks bias.
Intention-to-treat estimates the effect of being assigned the intervention under realistic conditions (effectiveness), which is usually the more relevant and conservative question for decision-making. Reviews often report both: agreement between them strengthens confidence, while a large divergence signals that adherence or missing data is shaping the result.
Practical challenges
Applying intention-to-treat strictly requires outcome data on everyone randomised, but in practice some participants are lost to follow-up. Handling this missing data is the central practical difficulty: simply analysing only those with complete data (a "complete-case" approach) can undermine the ITT principle. Statistical methods such as multiple imputation and pre-specified sensitivity analyses are used to address it. The CONSORT reporting standard asks trialists to state which analysis population was used and how missing data were handled, so readers can judge the robustness of the findings.
Key facts
At a glance
- Principle: Analyse participants as randomly assigned
- Preserves: The balance created by randomisation
- Estimates: Effectiveness under realistic conditions
- Contrast: Per-protocol analysis (adherent participants only)
- Challenge: Handling loss to follow-up and missing data
- Reported via: CONSORT analysis-population reporting
Common questions
FAQ
Why analyse participants who did not adhere to the intervention?+
Keeping every participant in their assigned group preserves the balance that randomisation created. Excluding non-adherent participants can reintroduce bias, because non-adherence is often linked to prognosis, so intention-to-treat keeps the randomised comparison intact.
What is the difference between intention-to-treat and per-protocol analysis?+
Intention-to-treat analyses everyone as randomised and estimates the effect of being assigned the intervention under realistic conditions. Per-protocol analyses only adherent participants and estimates the effect under ideal adherence, but by selecting on behaviour after randomisation it risks bias.
What is the main difficulty with intention-to-treat analysis?+
The main challenge is missing outcome data when participants are lost to follow-up, since strict ITT requires outcomes on everyone randomised. Researchers address this with methods such as multiple imputation and pre-specified sensitivity analyses, and report how missing data were handled.
Going deeper
Related on CASRAI
- Randomized controlled trial →
- Clinical trial →
- Selection bias →
- Double-blind study →
- Clinical research hub →
Sources
The step most authors miss
Doing CRediT right? Don’t stop at the statement.
A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.
Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.
