[Durgical](https://durgical.com/en/) › [DurgiPedia](https://durgical.com/en/durgipedia/) › [Epidemiology](https://durgical.com/en/epidemiology/) › [Study designs](https://durgical.com/en/epidemiology/areas/designs/) › Intention-to-treat analysis

# Intention-to-treat analysis

- Concept

Each patient is analysed in the group chance assigned them to, even if they never received or finished the treatment.

Updated Nov 5, 2026

## Key points

- ITT: everyone is analysed by assigned group: ‘as randomised, so analysed’.
- It preserves randomisation and, in a superiority trial, is usually conservative (pulls the effect towards zero).
- Per-protocol: only those who adhered; can exaggerate or create differences.
- Non-inferiority trials report both analyses, because ITT can falsely favour non-inferiority.
- Missing data complicate ITT and require assumptions (for example, worst case).

## What it is

In intention-to-treat (ITT) analysis every participant is counted in the group to which they were randomised, whether they adhered, dropped out or switched treatment. This keeps the balance that chance created and estimates the effect of offering the intervention in real practice. Per-protocol analysis includes only those who adhered, which can break that balance because non-adherent people tend to be different.

## Worked example

- 400 patients are randomised: 200 to surgery and 200 to medical treatment; outcome: death at 1 year.
- Of the 200 assigned to surgery, 30 are not operated on because they are too ill; 9 of them die.
- Among the 170 operated on, 17 die (10%); the medical group has 26 deaths out of 200 (13%).
- ITT: surgery arm = (17 + 9) / 200 = 26 / 200 = 13%, against 13%: no difference.
- Per protocol: 17 / 170 = 10% against 13%: surgery seems to lower mortality.
- The apparent benefit appears because the 30 sickest were removed from the surgical arm only.
- Reading: ITT keeps the groups comparable and shows that offering surgery did not change mortality.

## Intention to treat versus per protocol

| Feature | Intention to treat | Per protocol |
| --- | --- | --- |
| Who is included | Everyone randomised, in their assigned group | Only those who adhered to treatment |
| Preserves randomisation | Yes | Not necessarily |
| What it estimates | The effect of offering the intervention | The effect of receiving it in full |
| Typical bias | Can dilute the effect towards zero | Can exaggerate the effect |

## Markers

- ITT
- Per protocol

## In practice

- ITT answers the health system's question: what happens if this treatment is offered to patients like these?
- In trial risk-of-bias assessment, an analysis that drops people who deviated from their assigned treatment can raise the risk of bias.

## Common pitfalls

- Calling an analysis ITT when it excluded those who never received treatment (‘modified ITT’): check who was left out.
- Thinking ITT fixes missing data: without a patient's outcome, assumptions are still needed.

## Memory aid

Where chance put them, there they stay: no patient changes group in the analysis.

## Test yourself

Think of the answer, then tap the question to see it.

## What is intention-to-treat analysis?

Analysing each patient in the group to which they were randomised, whether or not they adhered to treatment.

## Why can per-protocol analysis be biased?

Because people who do not adhere tend to differ (sicker, more side effects), and excluding them breaks the balance chance created.

## In a superiority trial, which way does ITT usually push the effect estimate?

Towards zero: it is conservative because it includes people who did not receive the full intervention.

[Add to my cards (3)](https://durgical.app/estudiar/?add=/en/epidemiology/intention-to-treat-analysis/)

## Connections

## Related 5

- [Randomised controlled trial](https://durgical.com/en/epidemiology/randomised-controlled-trial/)
- [Attrition bias](https://durgical.com/en/epidemiology/attrition-bias/)
- [Superiority, equivalence and non-inferiority trials](https://durgical.com/en/epidemiology/superiority-equivalence-and-non-inferiority-trials/)
- [Randomisation and allocation concealment](https://durgical.com/en/epidemiology/randomisation-and-allocation-concealment/)
- [Selection bias](https://durgical.com/en/epidemiology/selection-bias/)

[← All of epidemiology](https://durgical.com/en/epidemiology/)

Educational content on epidemiological and statistical methods; the examples use made-up numbers to teach the calculation. Not a clinical or legal guide: current Colombian rules are in the official sources.

Sources: [Chapter 8: Assessing risk of bias in a randomized trial](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-08) (Cochrane, 2024); [Attrition bias | Catalog of Bias](https://catalogofbias.org/biases/attrition-bias/) (Catalogue of Bias Collaboration, 2017).

## **Spotted a mistake?** We are always improving for the community. Tell us what is wrong and our team will review it. Report a mistake

Representative model, for educational purposes only.

---
Source: https://durgical.com/en/epidemiology/intention-to-treat-analysis/
License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Cite as: DurgiPedia (durgical.com), https://durgical.com/en/epidemiology/intention-to-treat-analysis/