Skip to content
Learn Motion
ExploreHow it worksMembership
Log in
Learn Motion

How Doctors Judge Whether a Medical Study Can Be Trusted

1The Clinical Question and Why Study Design Follows From It2Randomization, Allocation, and the Logic of Comparison3Blinding, Follow-Up, and Who Actually Got Analyzed4Reading the Result: Effect Size, Uncertainty, and Significance5Applicability: Does This Result Fit My Patient?6Combining Studies and Forming a Verdict
Blinding, Follow-Up, and Who Actually Got Analyzed

Intention-to-Treat or Per-Protocol: Which Question Are You Asking?

3 / 3
The choice between intention-to-treat and per-protocol is a choice about which question the trial answers. Intention-to-treat keeps everyone in their assigned arm, so the groups stay exchangeable and the analysis estimates the effect of being assigned to the treatment strategy — which is usually the question a clinician faces, because real patients do not always adhere. Per-protocol includes only those who adhered, and that breaks the randomization: adherence is often related to prognosis, so the remaining groups are no longer exchangeable. Look at the worked example. A 40% benefit per-protocol that shrinks to 10% under intention-to-treat tells you the benefit is concentrated in the people who adhered. That could mean the treatment only works in adherent patients, or it could mean adherence selected a better-prognosis group. To tell these apart, ask why people deviated and whether that reason is related to the outcome. If people stopped because they were deteriorating, the per-protocol group is enriched with people who were doing well, and the 40% is inflated.
0:00 / 0:00

Intention-to-treat preserves randomization

ITT keeps every randomized participant in their assigned arm. Because assignment, not adherence, defined the groups, the groups remain exchangeable, and the analysis estimates the effect of being assigned to the treatment strategy.

Per-protocol breaks the randomization

Per-protocol analysis includes only those who adhered. Adherence is often related to prognosis — people who adhere may be healthier or more motivated — so the remaining groups are no longer exchangeable, and the apparent benefit can be an artifact of who adhered rather than an effect of the treatment.

A large per-protocol benefit that shrinks under ITT

Suppose a trial reports a 40% relative risk reduction in the per-protocol analysis but only 10% under ITT. The per-protocol benefit is concentrated in the people who adhered. That could mean the treatment only works in adherent patients, or it could mean that adherence selected a better-prognosis group. To tell these apart, look at why people deviated and whether the deviation was related to the outcome. If the people who stopped the treatment did so because they were deteriorating, the per-protocol group is enriched with people who were doing well, and the 40% is inflated.

The appraisal question

Neither analysis is universally correct. Ask whether the analysis population matches the question the trial claims to answer, and whether the reasons for exclusion are related to the outcome.

References

  1. [1]Cochrane Handbook for Systematic Reviews of Interventions, Chapter 8: Assessing risk of bias in a randomized trialtraining.cochrane.org
Previous3 / 3Next

Learn Motion

Generate a course. Learn it properly.

Operated by Wuhan Daoyin Technology Co., Ltd.

Contact: [email protected]
Privacy PolicyTerms of Service

© 2026 Learn Motion