Concept
Separate being offered a programme from completing it
Assignment and perfect attendance are not the same intervention question.
Before you use it
- Use when
- A trial's result is being interpreted without checking which analysis population it uses.
- Avoid when
- Do not assume every intention-to-treat label guarantees complete data, or every adherence analysis is invalid.
An idea to explore.
Try this idea.
Inspect the example, try the idea in your situation, then check your next move.
The idea
An assignment-based analysis asks about the effect of allocation. An adherence-focused analysis asks something different and needs methods that address selection. Read the actual inclusions, not only the label beside the table.
An example
A training provider removes everyone who missed sessions before comparing results. The remaining participants may differ in more than their exposure to the lessons.
Check your result
Can you explain who remained in each analysed group and why?
Keep this limit in mind
- Do not assume every intention-to-treat label guarantees complete data, or every adherence analysis is invalid.
My learning
Read. Recall. Repeat.
Save this idea and return to it on any device.
More from this collection
Check what a study can actually tell you
Keep evidence certainty attached to its outcomeCheck who could enter the sampleEvidence and sources (1)
Assignment effects and adherence effects are different questions; naive per-protocol exclusions can introduce bias.
Claim basis: technical_definition. Claim limit: The label intention-to-treat does not itself establish how missing data were handled. Source access: selected_html_passages; accessed on: 2026-09-17; inspected sections: ["8.2.2", "8.2.3 signalling questions", "8.3", "8.5", "8.6"]. Source note: The elements introduce questions for readers; they do not reproduce or claim to complete the RoB 2 assessment. Agent scope review, 2026-10-11: Cochrane 8.2.2/8.5 separate assignment and adherence questions. Naive exclusions can bias a comparison; the ITT label alone does not settle missing-data handling.
Chapter 8: Assessing risk of bias in a randomized trial · 8.2.2; 8.5
All sources (1)
Edition and review
Version 0.1.0 · Published
Agent review: Agent editorial and scoped evidence review · 2026-10-11