Posters · Glossary
What is intention to treat?
Intention to treat, or ITT, is a way of analysing a randomised trial in which every participant is counted in the group they were randomly assigned to, whether or not they took the treatment, stopped early or switched. It keeps the balance that randomisation created, so the comparison estimates the effect of offering a treatment in practice.
Every trial loses people along the way. How the analysis treats those people decides whether the result still means what randomisation promised.
Indunil Asanka · Co-founder
4 min read · Published
| Analysis | Who is included, and in which group | What it estimates | Main weakness |
|---|---|---|---|
| Intention to treat | Everyone randomised, in the group they were assigned to | The effect of being offered or assigned the treatment | Dilutes the effect when many do not adhere |
| Modified intention to treat | Everyone randomised who meets a stated condition, such as receiving one dose | Close to ITT, with defined exclusions | Exclusions can reintroduce bias if they depend on the arm |
| Per protocol | Only those who followed the protocol | The effect among adherent participants | Adherent people may differ from those who stopped |
| As treated | Everyone, grouped by what they actually received | The effect of the treatment as taken | Breaks randomisation; behaves like an observational study |
Why it matters
Randomisation makes the groups comparable at the start, in both the factors researchers measured and those they did not. If people who stop a treatment are then removed or moved, the groups being compared are no longer the ones randomisation created. People who drop out of an exercise program, for example, may be the ones in more pain. Keeping everyone in their assigned group preserves the fair comparison, at the cost of measuring the effect of a policy of treatment rather than of perfect adherence.
What the reporting guidelines ask
The CONSORT 2010 explanation and elaboration paper asks trials to report, for each group, the number of participants analysed and whether the analysis was by the original assigned groups, and it explains why analysis by assigned group is preferred. The Cochrane Handbook's risk of bias tool distinguishes the effect of assignment, which an intention to treat analysis estimates, from the effect of adhering to the intervention, which needs different methods.
Missing outcomes
A full intention to treat analysis needs an outcome for everyone randomised, which is rare. Trials handle the gap with methods such as multiple imputation, and say what they assumed about the missing people. A review of the concept in Perspectives in Clinical Research notes that many trials labelled intention to treat in fact exclude participants with missing data, so check the numbers in the flow diagram against the numbers analysed.
Reading it on a poster
Look for the number randomised in each arm and the number analysed, the label on the main analysis, and how missing data were handled. A per protocol result reported alongside is useful as a sensitivity analysis, and if the two agree the finding is more robust. If only a per protocol analysis is shown, the result may overstate the benefit, because it compares people who stayed with a treatment against the whole control group.
Showing it on a poster board
Ten of the 40 poster boards carry a CONSORT flow block, which shows how many people were assessed, allocated, followed up and analysed in each arm, the numbers that make an intention to treat claim checkable. Stat blocks, on 26 boards, suit the main effect with its interval and a second line for the per protocol result. The randomised trial example uses this layout; give every count in your brief, because the generator does not invent statistics.
Questions people ask
Does intention to treat underestimate a treatment's effect?
It can, when many participants do not take the treatment as assigned, because their outcomes dilute the difference between groups. That is deliberate: it reflects what happens when the treatment is offered in real practice. The per protocol estimate shows the effect among adherent people, but with a higher risk of bias.
Is intention to treat required in non inferiority trials?
Non inferiority trials usually report both intention to treat and per protocol analyses. Because dilution in an intention to treat analysis makes treatments look more similar, it can favour a conclusion of non inferiority, so reviewers look for consistent results from both before accepting the claim.
What is modified intention to treat?
A version that excludes participants who meet a pre specified condition, such as never receiving a single dose or turning out not to meet the entry criteria. It can be reasonable, but the term is used loosely. A good report states exactly who was excluded, how many from each arm, and why.
Can observational studies use intention to treat?
The term belongs to randomised trials, because it protects the balance created by random allocation. Observational studies that emulate a trial sometimes use an analogous approach, counting people by the treatment they started, but they still need to deal with confounding, which randomisation would otherwise have handled.
How should dropouts be shown on a poster?
In a participant flow diagram, with numbers lost to follow up and discontinued in each arm, and a reason where known. Then state how those participants were handled in the analysis, such as imputation. Readers use those numbers to judge whether losses could have changed the result.
Why do some trials report intention to treat and per protocol results together?
Because each answers a different question and has different weaknesses. When both point the same way, readers can be more confident the result is not an artefact of dropouts or poor adherence. When they diverge, the reasons need explaining, such as many people in one arm stopping treatment early.
Make one with posters
The button opens the generator with this use case already described. Change the wording to match your own.
Create a poster with OneCraftRelated questions
- What is an RCT (randomised controlled trial)?What is an RCT: a study that randomly allocates people to an intervention or a control so groups differ only by chance. How it limits bias, and what it costs.
- What is a pilot study?A pilot study runs a future trial, or part of it, on a small scale to test whether it can be done. Pilot against feasibility study, and progression criteria.
- What is number needed to treat (NNT)?What is number needed to treat (NNT): how many people need a treatment for one more to benefit. The formula from absolute risk reduction and a worked trial.
- What is a confounding variable?What is a confounding variable: a factor linked to both exposure and outcome that distorts their link. An example, how to control it, confounder vs mediator.
Step by step in the builder: Add charts and diagrams.
Written and checked by the OneCraft team. Last checked .